net: use skb->csum_not_inet to identify packets needing crc32c
[linux-block.git] / include / linux / skbuff.h
CommitLineData
1da177e4
LT
1/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
1da177e4 17#include <linux/kernel.h>
fe55f6d5 18#include <linux/kmemcheck.h>
1da177e4
LT
19#include <linux/compiler.h>
20#include <linux/time.h>
187f1882 21#include <linux/bug.h>
1da177e4 22#include <linux/cache.h>
56b17425 23#include <linux/rbtree.h>
51f3d02b 24#include <linux/socket.h>
1da177e4 25
60063497 26#include <linux/atomic.h>
1da177e4
LT
27#include <asm/types.h>
28#include <linux/spinlock.h>
1da177e4 29#include <linux/net.h>
3fc7e8a6 30#include <linux/textsearch.h>
1da177e4 31#include <net/checksum.h>
a80958f4 32#include <linux/rcupdate.h>
b7aa0bf7 33#include <linux/hrtimer.h>
131ea667 34#include <linux/dma-mapping.h>
c8f44aff 35#include <linux/netdev_features.h>
363ec392 36#include <linux/sched.h>
e6017571 37#include <linux/sched/clock.h>
1bd758eb 38#include <net/flow_dissector.h>
a60e3cc7 39#include <linux/splice.h>
72b31f72 40#include <linux/in6.h>
8b10cab6 41#include <linux/if_packet.h>
f70ea018 42#include <net/flow.h>
1da177e4 43
7a6ae71b
TH
44/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
78ea85f1
DB
89 *
90 * CHECKSUM_NONE:
91 *
7a6ae71b 92 * Device did not checksum this packet e.g. due to lack of capabilities.
78ea85f1
DB
93 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
77cffe23
TH
100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
7a6ae71b
TH
102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
77cffe23
TH
104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
112 *
113 * skb->csum_level indicates the number of consecutive checksums found in
114 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
115 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
116 * and a device is able to verify the checksums for UDP (possibly zero),
117 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
118 * two. If the device were only able to verify the UDP checksum and not
119 * GRE, either because it doesn't support GRE checksum of because GRE
120 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
121 * not considered in this case).
78ea85f1
DB
122 *
123 * CHECKSUM_COMPLETE:
124 *
125 * This is the most generic way. The device supplied checksum of the _whole_
126 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
127 * hardware doesn't need to parse L3/L4 headers to implement this.
128 *
129 * Note: Even if device supports only some protocols, but is able to produce
130 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
131 *
132 * CHECKSUM_PARTIAL:
133 *
6edec0e6
TH
134 * A checksum is set up to be offloaded to a device as described in the
135 * output description for CHECKSUM_PARTIAL. This may occur on a packet
78ea85f1 136 * received directly from another Linux OS, e.g., a virtualized Linux kernel
6edec0e6
TH
137 * on the same host, or it may be set in the input path in GRO or remote
138 * checksum offload. For the purposes of checksum verification, the checksum
139 * referred to by skb->csum_start + skb->csum_offset and any preceding
140 * checksums in the packet are considered verified. Any checksums in the
141 * packet that are after the checksum being offloaded are not considered to
142 * be verified.
78ea85f1 143 *
7a6ae71b
TH
144 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
145 * in the skb->ip_summed for a packet. Values are:
78ea85f1
DB
146 *
147 * CHECKSUM_PARTIAL:
148 *
7a6ae71b 149 * The driver is required to checksum the packet as seen by hard_start_xmit()
78ea85f1 150 * from skb->csum_start up to the end, and to record/write the checksum at
7a6ae71b
TH
151 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
152 * csum_start and csum_offset values are valid values given the length and
153 * offset of the packet, however they should not attempt to validate that the
154 * checksum refers to a legitimate transport layer checksum-- it is the
155 * purview of the stack to validate that csum_start and csum_offset are set
156 * correctly.
157 *
158 * When the stack requests checksum offload for a packet, the driver MUST
159 * ensure that the checksum is set correctly. A driver can either offload the
160 * checksum calculation to the device, or call skb_checksum_help (in the case
161 * that the device does not support offload for a particular checksum).
162 *
163 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
164 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
165 * checksum offload capability. If a device has limited checksum capabilities
166 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
167 * described above) a helper function can be called to resolve
168 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
169 * function takes a spec argument that describes the protocol layer that is
170 * supported for checksum offload and can be called for each packet. If a
171 * packet does not match the specification for offload, skb_checksum_help
172 * is called to resolve the checksum.
78ea85f1 173 *
7a6ae71b 174 * CHECKSUM_NONE:
78ea85f1 175 *
7a6ae71b
TH
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
78ea85f1
DB
178 *
179 * CHECKSUM_UNNECESSARY:
180 *
7a6ae71b
TH
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
187 *
188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
dba00306
DC
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
7a6ae71b
TH
199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
78ea85f1
DB
217 */
218
60476372 219/* Don't change this without changing skb_csum_unnecessary! */
78ea85f1
DB
220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
1da177e4 224
77cffe23
TH
225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
0bec8c88 228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
fc910a27 229#define SKB_WITH_OVERHEAD(X) \
deea84b0 230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
fc910a27
DM
231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
1da177e4
LT
233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
87fb4b7b
ED
236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
1da177e4 241struct net_device;
716ea3a7 242struct scatterlist;
9c55e01c 243struct pipe_inode_info;
a8f820aa 244struct iov_iter;
fd11a83d 245struct napi_struct;
1da177e4 246
5f79e0f9 247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
1da177e4
LT
248struct nf_conntrack {
249 atomic_t use;
1da177e4 250};
5f79e0f9 251#endif
1da177e4 252
34666d46 253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 254struct nf_bridge_info {
bf1ac5ca 255 atomic_t use;
3eaf4025
FW
256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
7fb48c5b 260 } orig_proto:8;
72b1e5e4
FW
261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
411ffb4f 264 __u16 frag_max_size;
bf1ac5ca 265 struct net_device *physindev;
63cdbc06
FW
266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
7fb48c5b 269 union {
72b1e5e4 270 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
72b1e5e4
FW
273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
72b31f72 279 };
1da177e4
LT
280};
281#endif
282
1da177e4
LT
283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
9d4dde52
IC
294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
a715dea3 300 */
9d4dde52 301#if (65536/PAGE_SIZE + 1) < 16
eec00954 302#define MAX_SKB_FRAGS 16UL
a715dea3 303#else
9d4dde52 304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 305#endif
5f74f82e 306extern int sysctl_max_skb_frags;
1da177e4 307
3953c46c
MRL
308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
1da177e4
LT
313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
a8605c60
IC
316 struct {
317 struct page *p;
318 } page;
cb4dfe56 319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
a309bb07
DM
320 __u32 page_offset;
321 __u32 size;
cb4dfe56
ED
322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
1da177e4
LT
326};
327
9e903e08
ED
328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
ac45f602
PO
348#define HAVE_HW_TIME_STAMP
349
350/**
d3a21be8 351 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
352 * @hwtstamp: hardware time stamp transformed into duration
353 * since arbitrary point in time
ac45f602
PO
354 *
355 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 356 * skb->tstamp.
ac45f602
PO
357 *
358 * hwtstamps can only be compared against other hwtstamps from
359 * the same device.
360 *
361 * This structure is attached to packets as part of the
362 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
363 */
364struct skb_shared_hwtstamps {
365 ktime_t hwtstamp;
ac45f602
PO
366};
367
2244d07b
OH
368/* Definitions for tx_flags in struct skb_shared_info */
369enum {
370 /* generate hardware time stamp */
371 SKBTX_HW_TSTAMP = 1 << 0,
372
e7fd2885 373 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
374 SKBTX_SW_TSTAMP = 1 << 1,
375
376 /* device driver is going to provide hardware time stamp */
377 SKBTX_IN_PROGRESS = 1 << 2,
378
a6686f2f 379 /* device driver supports TX zero-copy buffers */
62b1a8ab 380 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
381
382 /* generate wifi status information (where possible) */
62b1a8ab 383 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
384
385 /* This indicates at least one fragment might be overwritten
386 * (as in vmsplice(), sendfile() ...)
387 * If we need to compute a TX checksum, we'll need to copy
388 * all frags to avoid possible bad checksum
389 */
390 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
391
392 /* generate software time stamp when entering packet scheduling */
393 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
394};
395
e1c8a607 396#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 397 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
398#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
399
a6686f2f
SM
400/*
401 * The callback notifies userspace to release buffers when skb DMA is done in
402 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
403 * The zerocopy_success argument is true if zero copy transmit occurred,
404 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
405 * The ctx field is used to track device context.
406 * The desc field is used to track userspace buffer index.
a6686f2f
SM
407 */
408struct ubuf_info {
e19d6763 409 void (*callback)(struct ubuf_info *, bool zerocopy_success);
ca8f4fb2 410 void *ctx;
a6686f2f 411 unsigned long desc;
ac45f602
PO
412};
413
1da177e4
LT
414/* This data is invariant across clones and lives at
415 * the end of the header data, ie. at skb->end.
416 */
417struct skb_shared_info {
7f564528 418 unsigned short _unused;
9f42f126
IC
419 unsigned char nr_frags;
420 __u8 tx_flags;
7967168c
HX
421 unsigned short gso_size;
422 /* Warning: this field is not always filled in (UFO)! */
423 unsigned short gso_segs;
1da177e4 424 struct sk_buff *frag_list;
ac45f602 425 struct skb_shared_hwtstamps hwtstamps;
7f564528 426 unsigned int gso_type;
09c2d251 427 u32 tskey;
9f42f126 428 __be32 ip6_frag_id;
ec7d2f2c
ED
429
430 /*
431 * Warning : all fields before dataref are cleared in __alloc_skb()
432 */
433 atomic_t dataref;
434
69e3c75f
JB
435 /* Intermediate layers must ensure that destructor_arg
436 * remains valid until skb destructor */
437 void * destructor_arg;
a6686f2f 438
fed66381
ED
439 /* must be last field, see pskb_expand_head() */
440 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
441};
442
443/* We divide dataref into two halves. The higher 16 bits hold references
444 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
445 * the entire skb->data. A clone of a headerless skb holds the length of
446 * the header in skb->hdr_len.
1da177e4
LT
447 *
448 * All users must obey the rule that the skb->data reference count must be
449 * greater than or equal to the payload reference count.
450 *
451 * Holding a reference to the payload part means that the user does not
452 * care about modifications to the header part of skb->data.
453 */
454#define SKB_DATAREF_SHIFT 16
455#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
456
d179cd12
DM
457
458enum {
c8753d55
VS
459 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
460 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
461 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
462};
463
7967168c
HX
464enum {
465 SKB_GSO_TCPV4 = 1 << 0,
f83ef8c0 466 SKB_GSO_UDP = 1 << 1,
576a30eb
HX
467
468 /* This indicates the skb is from an untrusted source. */
469 SKB_GSO_DODGY = 1 << 2,
b0da8537
MC
470
471 /* This indicates the tcp segment has CWR set. */
f83ef8c0
HX
472 SKB_GSO_TCP_ECN = 1 << 3,
473
cbc53e08 474 SKB_GSO_TCP_FIXEDID = 1 << 4,
01d5b2fc 475
cbc53e08 476 SKB_GSO_TCPV6 = 1 << 5,
68c33163 477
cbc53e08 478 SKB_GSO_FCOE = 1 << 6,
73136267 479
cbc53e08 480 SKB_GSO_GRE = 1 << 7,
0d89d203 481
cbc53e08 482 SKB_GSO_GRE_CSUM = 1 << 8,
cb32f511 483
7e13318d 484 SKB_GSO_IPXIP4 = 1 << 9,
61c1db7f 485
7e13318d 486 SKB_GSO_IPXIP6 = 1 << 10,
0f4f4ffa 487
cbc53e08 488 SKB_GSO_UDP_TUNNEL = 1 << 11,
4749c09c 489
cbc53e08
AD
490 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
491
802ab55a
AD
492 SKB_GSO_PARTIAL = 1 << 13,
493
494 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
90017acc
MRL
495
496 SKB_GSO_SCTP = 1 << 15,
c7ef8f0c
SK
497
498 SKB_GSO_ESP = 1 << 16,
7967168c
HX
499};
500
2e07fa9c
ACM
501#if BITS_PER_LONG > 32
502#define NET_SKBUFF_DATA_USES_OFFSET 1
503#endif
504
505#ifdef NET_SKBUFF_DATA_USES_OFFSET
506typedef unsigned int sk_buff_data_t;
507#else
508typedef unsigned char *sk_buff_data_t;
509#endif
510
1da177e4
LT
511/**
512 * struct sk_buff - socket buffer
513 * @next: Next buffer in list
514 * @prev: Previous buffer in list
363ec392 515 * @tstamp: Time we arrived/left
56b17425 516 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 517 * @sk: Socket we are owned by
1da177e4 518 * @dev: Device we arrived on/are leaving by
d84e0bd7 519 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 520 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 521 * @sp: the security path, used for xfrm
1da177e4
LT
522 * @len: Length of actual data
523 * @data_len: Data length
524 * @mac_len: Length of link layer header
334a8132 525 * @hdr_len: writable header length of cloned skb
663ead3b
HX
526 * @csum: Checksum (must include start/offset pair)
527 * @csum_start: Offset from skb->head where checksumming should start
528 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 529 * @priority: Packet queueing priority
60ff7467 530 * @ignore_df: allow local fragmentation
1da177e4 531 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 532 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
533 * @nohdr: Payload reference only, must not modify header
534 * @pkt_type: Packet class
c83c2486 535 * @fclone: skbuff clone status
c83c2486 536 * @ipvs_property: skbuff is owned by ipvs
e7246e12 537 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 538 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
539 * @tc_redirected: packet was redirected by a tc action
540 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
541 * @peeked: this packet has been seen already, so stats have been
542 * done for it, don't do them again
ba9dda3a 543 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
544 * @protocol: Packet protocol from driver
545 * @destructor: Destruct function
a9e419dc 546 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 547 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 548 * @skb_iif: ifindex of device we arrived on
1da177e4 549 * @tc_index: Traffic control index
61b905da 550 * @hash: the packet hash
d84e0bd7 551 * @queue_mapping: Queue mapping for multiqueue devices
0b725a2c 552 * @xmit_more: More SKBs are pending for this queue
553a5672 553 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 554 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 555 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 556 * ports.
a3b18ddb 557 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
558 * @wifi_acked_valid: wifi_acked was set
559 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 560 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
dba00306 561 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
4ff06203 562 * @dst_pending_confirm: need to confirm neighbour
06021292 563 * @napi_id: id of the NAPI struct this skb came from
984bc16c 564 * @secmark: security marking
d84e0bd7 565 * @mark: Generic packet mark
86a9bad3 566 * @vlan_proto: vlan encapsulation protocol
6aa895b0 567 * @vlan_tci: vlan tag control information
0d89d203 568 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
569 * @inner_transport_header: Inner transport layer header (encapsulation)
570 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 571 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
572 * @transport_header: Transport layer header
573 * @network_header: Network layer header
574 * @mac_header: Link layer header
575 * @tail: Tail pointer
576 * @end: End pointer
577 * @head: Head of buffer
578 * @data: Data head pointer
579 * @truesize: Buffer size
580 * @users: User count - see {datagram,tcp}.c
1da177e4
LT
581 */
582
583struct sk_buff {
363ec392 584 union {
56b17425
ED
585 struct {
586 /* These two members must be first. */
587 struct sk_buff *next;
588 struct sk_buff *prev;
589
590 union {
591 ktime_t tstamp;
9a568de4 592 u64 skb_mstamp;
56b17425
ED
593 };
594 };
595 struct rb_node rbnode; /* used in netem & tcp stack */
363ec392 596 };
da3f5cf1 597 struct sock *sk;
1da177e4 598
c84d9490
ED
599 union {
600 struct net_device *dev;
601 /* Some protocols might use this space to store information,
602 * while device pointer would be NULL.
603 * UDP receive path is one user.
604 */
605 unsigned long dev_scratch;
606 };
1da177e4
LT
607 /*
608 * This is the control buffer. It is free to use for every
609 * layer. Please put your private variables there. If you
610 * want to keep them across layers you have to do a skb_clone()
611 * first. This is owned by whoever has the skb queued ATM.
612 */
da3f5cf1 613 char cb[48] __aligned(8);
1da177e4 614
7fee226a 615 unsigned long _skb_refdst;
b1937227 616 void (*destructor)(struct sk_buff *skb);
da3f5cf1
FF
617#ifdef CONFIG_XFRM
618 struct sec_path *sp;
b1937227
ED
619#endif
620#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 621 unsigned long _nfct;
b1937227 622#endif
85224844 623#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
b1937227 624 struct nf_bridge_info *nf_bridge;
da3f5cf1 625#endif
1da177e4 626 unsigned int len,
334a8132
PM
627 data_len;
628 __u16 mac_len,
629 hdr_len;
b1937227
ED
630
631 /* Following fields are _not_ copied in __copy_skb_header()
632 * Note that queue_mapping is here mostly to fill a hole.
633 */
fe55f6d5 634 kmemcheck_bitfield_begin(flags1);
b1937227 635 __u16 queue_mapping;
36bbef52
DB
636
637/* if you move cloned around you also must adapt those constants */
638#ifdef __BIG_ENDIAN_BITFIELD
639#define CLONED_MASK (1 << 7)
640#else
641#define CLONED_MASK 1
642#endif
643#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
644
645 __u8 __cloned_offset[0];
b1937227 646 __u8 cloned:1,
6869c4d8 647 nohdr:1,
b84f4cc9 648 fclone:2,
a59322be 649 peeked:1,
b1937227 650 head_frag:1,
36bbef52
DB
651 xmit_more:1,
652 __unused:1; /* one bit hole */
fe55f6d5 653 kmemcheck_bitfield_end(flags1);
4031ae6e 654
b1937227
ED
655 /* fields enclosed in headers_start/headers_end are copied
656 * using a single memcpy() in __copy_skb_header()
657 */
ebcf34f3 658 /* private: */
b1937227 659 __u32 headers_start[0];
ebcf34f3 660 /* public: */
4031ae6e 661
233577a2
HFS
662/* if you move pkt_type around you also must adapt those constants */
663#ifdef __BIG_ENDIAN_BITFIELD
664#define PKT_TYPE_MAX (7 << 5)
665#else
666#define PKT_TYPE_MAX 7
1da177e4 667#endif
233577a2 668#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 669
233577a2 670 __u8 __pkt_type_offset[0];
b1937227 671 __u8 pkt_type:3;
c93bdd0e 672 __u8 pfmemalloc:1;
b1937227 673 __u8 ignore_df:1;
b1937227
ED
674
675 __u8 nf_trace:1;
676 __u8 ip_summed:2;
3853b584 677 __u8 ooo_okay:1;
61b905da 678 __u8 l4_hash:1;
a3b18ddb 679 __u8 sw_hash:1;
6e3e939f
JB
680 __u8 wifi_acked_valid:1;
681 __u8 wifi_acked:1;
b1937227 682
3bdc0eba 683 __u8 no_fcs:1;
77cffe23 684 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 685 __u8 encapsulation:1;
7e2b10c1 686 __u8 encap_hdr_csum:1;
5d0c2b95 687 __u8 csum_valid:1;
7e3cead5 688 __u8 csum_complete_sw:1;
b1937227 689 __u8 csum_level:2;
dba00306 690 __u8 csum_not_inet:1;
fe55f6d5 691
4ff06203 692 __u8 dst_pending_confirm:1;
b1937227
ED
693#ifdef CONFIG_IPV6_NDISC_NODETYPE
694 __u8 ndisc_nodetype:2;
695#endif
696 __u8 ipvs_property:1;
8bce6d7d 697 __u8 inner_protocol_type:1;
e585f236 698 __u8 remcsum_offload:1;
6bc506b4
IS
699#ifdef CONFIG_NET_SWITCHDEV
700 __u8 offload_fwd_mark:1;
701#endif
e7246e12
WB
702#ifdef CONFIG_NET_CLS_ACT
703 __u8 tc_skip_classify:1;
8dc07fdb 704 __u8 tc_at_ingress:1;
bc31c905
WB
705 __u8 tc_redirected:1;
706 __u8 tc_from_ingress:1;
e7246e12 707#endif
b1937227
ED
708
709#ifdef CONFIG_NET_SCHED
710 __u16 tc_index; /* traffic control index */
b1937227 711#endif
fe55f6d5 712
b1937227
ED
713 union {
714 __wsum csum;
715 struct {
716 __u16 csum_start;
717 __u16 csum_offset;
718 };
719 };
720 __u32 priority;
721 int skb_iif;
722 __u32 hash;
723 __be16 vlan_proto;
724 __u16 vlan_tci;
2bd82484
ED
725#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
726 union {
727 unsigned int napi_id;
728 unsigned int sender_cpu;
729 };
97fc2f08 730#endif
984bc16c 731#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 732 __u32 secmark;
0c4f691f 733#endif
0c4f691f 734
3b885787
NH
735 union {
736 __u32 mark;
16fad69c 737 __u32 reserved_tailroom;
3b885787 738 };
1da177e4 739
8bce6d7d
TH
740 union {
741 __be16 inner_protocol;
742 __u8 inner_ipproto;
743 };
744
1a37e412
SH
745 __u16 inner_transport_header;
746 __u16 inner_network_header;
747 __u16 inner_mac_header;
b1937227
ED
748
749 __be16 protocol;
1a37e412
SH
750 __u16 transport_header;
751 __u16 network_header;
752 __u16 mac_header;
b1937227 753
ebcf34f3 754 /* private: */
b1937227 755 __u32 headers_end[0];
ebcf34f3 756 /* public: */
b1937227 757
1da177e4 758 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 759 sk_buff_data_t tail;
4305b541 760 sk_buff_data_t end;
1da177e4 761 unsigned char *head,
4305b541 762 *data;
27a884dc
ACM
763 unsigned int truesize;
764 atomic_t users;
1da177e4
LT
765};
766
767#ifdef __KERNEL__
768/*
769 * Handling routines are only of interest to the kernel
770 */
771#include <linux/slab.h>
772
1da177e4 773
c93bdd0e
MG
774#define SKB_ALLOC_FCLONE 0x01
775#define SKB_ALLOC_RX 0x02
fd11a83d 776#define SKB_ALLOC_NAPI 0x04
c93bdd0e
MG
777
778/* Returns true if the skb was allocated from PFMEMALLOC reserves */
779static inline bool skb_pfmemalloc(const struct sk_buff *skb)
780{
781 return unlikely(skb->pfmemalloc);
782}
783
7fee226a
ED
784/*
785 * skb might have a dst pointer attached, refcounted or not.
786 * _skb_refdst low order bit is set if refcount was _not_ taken
787 */
788#define SKB_DST_NOREF 1UL
789#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
790
a9e419dc 791#define SKB_NFCT_PTRMASK ~(7UL)
7fee226a
ED
792/**
793 * skb_dst - returns skb dst_entry
794 * @skb: buffer
795 *
796 * Returns skb dst_entry, regardless of reference taken or not.
797 */
adf30907
ED
798static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
799{
7fee226a
ED
800 /* If refdst was not refcounted, check we still are in a
801 * rcu_read_lock section
802 */
803 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
804 !rcu_read_lock_held() &&
805 !rcu_read_lock_bh_held());
806 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
807}
808
7fee226a
ED
809/**
810 * skb_dst_set - sets skb dst
811 * @skb: buffer
812 * @dst: dst entry
813 *
814 * Sets skb dst, assuming a reference was taken on dst and should
815 * be released by skb_dst_drop()
816 */
adf30907
ED
817static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
818{
7fee226a
ED
819 skb->_skb_refdst = (unsigned long)dst;
820}
821
932bc4d7
JA
822/**
823 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
824 * @skb: buffer
825 * @dst: dst entry
826 *
827 * Sets skb dst, assuming a reference was not taken on dst.
828 * If dst entry is cached, we do not take reference and dst_release
829 * will be avoided by refdst_drop. If dst entry is not cached, we take
830 * reference, so that last dst_release can destroy the dst immediately.
831 */
832static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
833{
dbfc4fb7
HFS
834 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
835 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 836}
7fee226a
ED
837
838/**
25985edc 839 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
840 * @skb: buffer
841 */
842static inline bool skb_dst_is_noref(const struct sk_buff *skb)
843{
844 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
845}
846
511c3f92
ED
847static inline struct rtable *skb_rtable(const struct sk_buff *skb)
848{
adf30907 849 return (struct rtable *)skb_dst(skb);
511c3f92
ED
850}
851
8b10cab6
JHS
852/* For mangling skb->pkt_type from user space side from applications
853 * such as nft, tc, etc, we only allow a conservative subset of
854 * possible pkt_types to be set.
855*/
856static inline bool skb_pkt_type_ok(u32 ptype)
857{
858 return ptype <= PACKET_OTHERHOST;
859}
860
7965bd4d
JP
861void kfree_skb(struct sk_buff *skb);
862void kfree_skb_list(struct sk_buff *segs);
863void skb_tx_error(struct sk_buff *skb);
864void consume_skb(struct sk_buff *skb);
865void __kfree_skb(struct sk_buff *skb);
d7e8883c 866extern struct kmem_cache *skbuff_head_cache;
bad43ca8 867
7965bd4d
JP
868void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
869bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
870 bool *fragstolen, int *delta_truesize);
bad43ca8 871
7965bd4d
JP
872struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
873 int node);
2ea2f62c 874struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 875struct sk_buff *build_skb(void *data, unsigned int frag_size);
d179cd12 876static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 877 gfp_t priority)
d179cd12 878{
564824b0 879 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
880}
881
2e4e4410
ED
882struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
883 unsigned long data_len,
884 int max_page_order,
885 int *errcode,
886 gfp_t gfp_mask);
887
d0bf4a9e
ED
888/* Layout of fast clones : [skb1][skb2][fclone_ref] */
889struct sk_buff_fclones {
890 struct sk_buff skb1;
891
892 struct sk_buff skb2;
893
894 atomic_t fclone_ref;
895};
896
897/**
898 * skb_fclone_busy - check if fclone is busy
293de7de 899 * @sk: socket
d0bf4a9e
ED
900 * @skb: buffer
901 *
bda13fed 902 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
903 * Some drivers call skb_orphan() in their ndo_start_xmit(),
904 * so we also check that this didnt happen.
d0bf4a9e 905 */
39bb5e62
ED
906static inline bool skb_fclone_busy(const struct sock *sk,
907 const struct sk_buff *skb)
d0bf4a9e
ED
908{
909 const struct sk_buff_fclones *fclones;
910
911 fclones = container_of(skb, struct sk_buff_fclones, skb1);
912
913 return skb->fclone == SKB_FCLONE_ORIG &&
6ffe75eb 914 atomic_read(&fclones->fclone_ref) > 1 &&
39bb5e62 915 fclones->skb2.sk == sk;
d0bf4a9e
ED
916}
917
d179cd12 918static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 919 gfp_t priority)
d179cd12 920{
c93bdd0e 921 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
922}
923
7965bd4d 924struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
0ebd0ac5
PM
925static inline struct sk_buff *alloc_skb_head(gfp_t priority)
926{
927 return __alloc_skb_head(priority, -1);
928}
929
7965bd4d
JP
930struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
931int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
932struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
933struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
934struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
935 gfp_t gfp_mask, bool fclone);
936static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
937 gfp_t gfp_mask)
938{
939 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
940}
7965bd4d
JP
941
942int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
943struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
944 unsigned int headroom);
945struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
946 int newtailroom, gfp_t priority);
25a91d8d
FD
947int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
948 int offset, int len);
7965bd4d
JP
949int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
950 int len);
951int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
952int skb_pad(struct sk_buff *skb, int pad);
ead2ceb0 953#define dev_kfree_skb(a) consume_skb(a)
1da177e4 954
7965bd4d
JP
955int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
956 int getfrag(void *from, char *to, int offset,
957 int len, int odd, struct sk_buff *skb),
958 void *from, int length);
e89e9cf5 959
be12a1fe
HFS
960int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
961 int offset, size_t size);
962
d94d9fee 963struct skb_seq_state {
677e90ed
TG
964 __u32 lower_offset;
965 __u32 upper_offset;
966 __u32 frag_idx;
967 __u32 stepped_offset;
968 struct sk_buff *root_skb;
969 struct sk_buff *cur_skb;
970 __u8 *frag_data;
971};
972
7965bd4d
JP
973void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
974 unsigned int to, struct skb_seq_state *st);
975unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
976 struct skb_seq_state *st);
977void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 978
7965bd4d 979unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 980 unsigned int to, struct ts_config *config);
3fc7e8a6 981
09323cc4
TH
982/*
983 * Packet hash types specify the type of hash in skb_set_hash.
984 *
985 * Hash types refer to the protocol layer addresses which are used to
986 * construct a packet's hash. The hashes are used to differentiate or identify
987 * flows of the protocol layer for the hash type. Hash types are either
988 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
989 *
990 * Properties of hashes:
991 *
992 * 1) Two packets in different flows have different hash values
993 * 2) Two packets in the same flow should have the same hash value
994 *
995 * A hash at a higher layer is considered to be more specific. A driver should
996 * set the most specific hash possible.
997 *
998 * A driver cannot indicate a more specific hash than the layer at which a hash
999 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1000 *
1001 * A driver may indicate a hash level which is less specific than the
1002 * actual layer the hash was computed on. For instance, a hash computed
1003 * at L4 may be considered an L3 hash. This should only be done if the
1004 * driver can't unambiguously determine that the HW computed the hash at
1005 * the higher layer. Note that the "should" in the second property above
1006 * permits this.
1007 */
1008enum pkt_hash_types {
1009 PKT_HASH_TYPE_NONE, /* Undefined type */
1010 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1011 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1012 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1013};
1014
bcc83839 1015static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1016{
bcc83839 1017 skb->hash = 0;
a3b18ddb 1018 skb->sw_hash = 0;
bcc83839
TH
1019 skb->l4_hash = 0;
1020}
1021
1022static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1023{
1024 if (!skb->l4_hash)
1025 skb_clear_hash(skb);
1026}
1027
1028static inline void
1029__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1030{
1031 skb->l4_hash = is_l4;
1032 skb->sw_hash = is_sw;
61b905da 1033 skb->hash = hash;
09323cc4
TH
1034}
1035
bcc83839
TH
1036static inline void
1037skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1038{
1039 /* Used by drivers to set hash from HW */
1040 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1041}
1042
1043static inline void
1044__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1045{
1046 __skb_set_hash(skb, hash, true, is_l4);
1047}
1048
e5276937 1049void __skb_get_hash(struct sk_buff *skb);
b917783c 1050u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1051u32 skb_get_poff(const struct sk_buff *skb);
1052u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1053 const struct flow_keys *keys, int hlen);
1054__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1055 void *data, int hlen_proto);
1056
1057static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1058 int thoff, u8 ip_proto)
1059{
1060 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1061}
1062
1063void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1064 const struct flow_dissector_key *key,
1065 unsigned int key_count);
1066
1067bool __skb_flow_dissect(const struct sk_buff *skb,
1068 struct flow_dissector *flow_dissector,
1069 void *target_container,
cd79a238
TH
1070 void *data, __be16 proto, int nhoff, int hlen,
1071 unsigned int flags);
e5276937
TH
1072
1073static inline bool skb_flow_dissect(const struct sk_buff *skb,
1074 struct flow_dissector *flow_dissector,
cd79a238 1075 void *target_container, unsigned int flags)
e5276937
TH
1076{
1077 return __skb_flow_dissect(skb, flow_dissector, target_container,
cd79a238 1078 NULL, 0, 0, 0, flags);
e5276937
TH
1079}
1080
1081static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1082 struct flow_keys *flow,
1083 unsigned int flags)
e5276937
TH
1084{
1085 memset(flow, 0, sizeof(*flow));
1086 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
cd79a238 1087 NULL, 0, 0, 0, flags);
e5276937
TH
1088}
1089
1090static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1091 void *data, __be16 proto,
cd79a238
TH
1092 int nhoff, int hlen,
1093 unsigned int flags)
e5276937
TH
1094{
1095 memset(flow, 0, sizeof(*flow));
1096 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
cd79a238 1097 data, proto, nhoff, hlen, flags);
e5276937
TH
1098}
1099
3958afa1 1100static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1101{
a3b18ddb 1102 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1103 __skb_get_hash(skb);
bfb564e7 1104
61b905da 1105 return skb->hash;
bfb564e7
KK
1106}
1107
20a17bf6 1108__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
f70ea018 1109
20a17bf6 1110static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1111{
c6cc1ca7
TH
1112 if (!skb->l4_hash && !skb->sw_hash) {
1113 struct flow_keys keys;
de4c1f8b 1114 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1115
de4c1f8b 1116 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1117 }
f70ea018
TH
1118
1119 return skb->hash;
1120}
1121
20a17bf6 1122__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
f70ea018 1123
20a17bf6 1124static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
f70ea018 1125{
c6cc1ca7
TH
1126 if (!skb->l4_hash && !skb->sw_hash) {
1127 struct flow_keys keys;
de4c1f8b 1128 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
c6cc1ca7 1129
de4c1f8b 1130 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1131 }
f70ea018
TH
1132
1133 return skb->hash;
1134}
1135
50fb7992
TH
1136__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1137
57bdf7f4
TH
1138static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1139{
61b905da 1140 return skb->hash;
57bdf7f4
TH
1141}
1142
3df7a74e
TH
1143static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1144{
61b905da 1145 to->hash = from->hash;
a3b18ddb 1146 to->sw_hash = from->sw_hash;
61b905da 1147 to->l4_hash = from->l4_hash;
3df7a74e
TH
1148};
1149
4305b541
ACM
1150#ifdef NET_SKBUFF_DATA_USES_OFFSET
1151static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1152{
1153 return skb->head + skb->end;
1154}
ec47ea82
AD
1155
1156static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1157{
1158 return skb->end;
1159}
4305b541
ACM
1160#else
1161static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1162{
1163 return skb->end;
1164}
ec47ea82
AD
1165
1166static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1167{
1168 return skb->end - skb->head;
1169}
4305b541
ACM
1170#endif
1171
1da177e4 1172/* Internal */
4305b541 1173#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1174
ac45f602
PO
1175static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1176{
1177 return &skb_shinfo(skb)->hwtstamps;
1178}
1179
1da177e4
LT
1180/**
1181 * skb_queue_empty - check if a queue is empty
1182 * @list: queue head
1183 *
1184 * Returns true if the queue is empty, false otherwise.
1185 */
1186static inline int skb_queue_empty(const struct sk_buff_head *list)
1187{
fd44b93c 1188 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1189}
1190
fc7ebb21
DM
1191/**
1192 * skb_queue_is_last - check if skb is the last entry in the queue
1193 * @list: queue head
1194 * @skb: buffer
1195 *
1196 * Returns true if @skb is the last buffer on the list.
1197 */
1198static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1199 const struct sk_buff *skb)
1200{
fd44b93c 1201 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1202}
1203
832d11c5
IJ
1204/**
1205 * skb_queue_is_first - check if skb is the first entry in the queue
1206 * @list: queue head
1207 * @skb: buffer
1208 *
1209 * Returns true if @skb is the first buffer on the list.
1210 */
1211static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1212 const struct sk_buff *skb)
1213{
fd44b93c 1214 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1215}
1216
249c8b42
DM
1217/**
1218 * skb_queue_next - return the next packet in the queue
1219 * @list: queue head
1220 * @skb: current buffer
1221 *
1222 * Return the next packet in @list after @skb. It is only valid to
1223 * call this if skb_queue_is_last() evaluates to false.
1224 */
1225static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1226 const struct sk_buff *skb)
1227{
1228 /* This BUG_ON may seem severe, but if we just return then we
1229 * are going to dereference garbage.
1230 */
1231 BUG_ON(skb_queue_is_last(list, skb));
1232 return skb->next;
1233}
1234
832d11c5
IJ
1235/**
1236 * skb_queue_prev - return the prev packet in the queue
1237 * @list: queue head
1238 * @skb: current buffer
1239 *
1240 * Return the prev packet in @list before @skb. It is only valid to
1241 * call this if skb_queue_is_first() evaluates to false.
1242 */
1243static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1244 const struct sk_buff *skb)
1245{
1246 /* This BUG_ON may seem severe, but if we just return then we
1247 * are going to dereference garbage.
1248 */
1249 BUG_ON(skb_queue_is_first(list, skb));
1250 return skb->prev;
1251}
1252
1da177e4
LT
1253/**
1254 * skb_get - reference buffer
1255 * @skb: buffer to reference
1256 *
1257 * Makes another reference to a socket buffer and returns a pointer
1258 * to the buffer.
1259 */
1260static inline struct sk_buff *skb_get(struct sk_buff *skb)
1261{
1262 atomic_inc(&skb->users);
1263 return skb;
1264}
1265
1266/*
1267 * If users == 1, we are the only owner and are can avoid redundant
1268 * atomic change.
1269 */
1270
1da177e4
LT
1271/**
1272 * skb_cloned - is the buffer a clone
1273 * @skb: buffer to check
1274 *
1275 * Returns true if the buffer was generated with skb_clone() and is
1276 * one of multiple shared copies of the buffer. Cloned buffers are
1277 * shared data so must not be written to under normal circumstances.
1278 */
1279static inline int skb_cloned(const struct sk_buff *skb)
1280{
1281 return skb->cloned &&
1282 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1283}
1284
14bbd6a5
PS
1285static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1286{
d0164adc 1287 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1288
1289 if (skb_cloned(skb))
1290 return pskb_expand_head(skb, 0, 0, pri);
1291
1292 return 0;
1293}
1294
1da177e4
LT
1295/**
1296 * skb_header_cloned - is the header a clone
1297 * @skb: buffer to check
1298 *
1299 * Returns true if modifying the header part of the buffer requires
1300 * the data to be copied.
1301 */
1302static inline int skb_header_cloned(const struct sk_buff *skb)
1303{
1304 int dataref;
1305
1306 if (!skb->cloned)
1307 return 0;
1308
1309 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1310 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1311 return dataref != 1;
1312}
1313
9580bf2e
ED
1314static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1315{
1316 might_sleep_if(gfpflags_allow_blocking(pri));
1317
1318 if (skb_header_cloned(skb))
1319 return pskb_expand_head(skb, 0, 0, pri);
1320
1321 return 0;
1322}
1323
1da177e4
LT
1324/**
1325 * skb_header_release - release reference to header
1326 * @skb: buffer to operate on
1327 *
1328 * Drop a reference to the header part of the buffer. This is done
1329 * by acquiring a payload reference. You must not read from the header
1330 * part of skb->data after this.
f4a775d1 1331 * Note : Check if you can use __skb_header_release() instead.
1da177e4
LT
1332 */
1333static inline void skb_header_release(struct sk_buff *skb)
1334{
1335 BUG_ON(skb->nohdr);
1336 skb->nohdr = 1;
1337 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1338}
1339
f4a775d1
ED
1340/**
1341 * __skb_header_release - release reference to header
1342 * @skb: buffer to operate on
1343 *
1344 * Variant of skb_header_release() assuming skb is private to caller.
1345 * We can avoid one atomic operation.
1346 */
1347static inline void __skb_header_release(struct sk_buff *skb)
1348{
1349 skb->nohdr = 1;
1350 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1351}
1352
1353
1da177e4
LT
1354/**
1355 * skb_shared - is the buffer shared
1356 * @skb: buffer to check
1357 *
1358 * Returns true if more than one person has a reference to this
1359 * buffer.
1360 */
1361static inline int skb_shared(const struct sk_buff *skb)
1362{
1363 return atomic_read(&skb->users) != 1;
1364}
1365
1366/**
1367 * skb_share_check - check if buffer is shared and if so clone it
1368 * @skb: buffer to check
1369 * @pri: priority for memory allocation
1370 *
1371 * If the buffer is shared the buffer is cloned and the old copy
1372 * drops a reference. A new clone with a single reference is returned.
1373 * If the buffer is not shared the original buffer is returned. When
1374 * being called from interrupt status or with spinlocks held pri must
1375 * be GFP_ATOMIC.
1376 *
1377 * NULL is returned on a memory allocation failure.
1378 */
47061bc4 1379static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1380{
d0164adc 1381 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1382 if (skb_shared(skb)) {
1383 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1384
1385 if (likely(nskb))
1386 consume_skb(skb);
1387 else
1388 kfree_skb(skb);
1da177e4
LT
1389 skb = nskb;
1390 }
1391 return skb;
1392}
1393
1394/*
1395 * Copy shared buffers into a new sk_buff. We effectively do COW on
1396 * packets to handle cases where we have a local reader and forward
1397 * and a couple of other messy ones. The normal one is tcpdumping
1398 * a packet thats being forwarded.
1399 */
1400
1401/**
1402 * skb_unshare - make a copy of a shared buffer
1403 * @skb: buffer to check
1404 * @pri: priority for memory allocation
1405 *
1406 * If the socket buffer is a clone then this function creates a new
1407 * copy of the data, drops a reference count on the old copy and returns
1408 * the new copy with the reference count at 1. If the buffer is not a clone
1409 * the original buffer is returned. When called with a spinlock held or
1410 * from interrupt state @pri must be %GFP_ATOMIC
1411 *
1412 * %NULL is returned on a memory allocation failure.
1413 */
e2bf521d 1414static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1415 gfp_t pri)
1da177e4 1416{
d0164adc 1417 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1418 if (skb_cloned(skb)) {
1419 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1420
1421 /* Free our shared copy */
1422 if (likely(nskb))
1423 consume_skb(skb);
1424 else
1425 kfree_skb(skb);
1da177e4
LT
1426 skb = nskb;
1427 }
1428 return skb;
1429}
1430
1431/**
1a5778aa 1432 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1433 * @list_: list to peek at
1434 *
1435 * Peek an &sk_buff. Unlike most other operations you _MUST_
1436 * be careful with this one. A peek leaves the buffer on the
1437 * list and someone else may run off with it. You must hold
1438 * the appropriate locks or have a private queue to do this.
1439 *
1440 * Returns %NULL for an empty list or a pointer to the head element.
1441 * The reference count is not incremented and the reference is therefore
1442 * volatile. Use with caution.
1443 */
05bdd2f1 1444static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1445{
18d07000
ED
1446 struct sk_buff *skb = list_->next;
1447
1448 if (skb == (struct sk_buff *)list_)
1449 skb = NULL;
1450 return skb;
1da177e4
LT
1451}
1452
da5ef6e5
PE
1453/**
1454 * skb_peek_next - peek skb following the given one from a queue
1455 * @skb: skb to start from
1456 * @list_: list to peek at
1457 *
1458 * Returns %NULL when the end of the list is met or a pointer to the
1459 * next element. The reference count is not incremented and the
1460 * reference is therefore volatile. Use with caution.
1461 */
1462static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1463 const struct sk_buff_head *list_)
1464{
1465 struct sk_buff *next = skb->next;
18d07000 1466
da5ef6e5
PE
1467 if (next == (struct sk_buff *)list_)
1468 next = NULL;
1469 return next;
1470}
1471
1da177e4 1472/**
1a5778aa 1473 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1474 * @list_: list to peek at
1475 *
1476 * Peek an &sk_buff. Unlike most other operations you _MUST_
1477 * be careful with this one. A peek leaves the buffer on the
1478 * list and someone else may run off with it. You must hold
1479 * the appropriate locks or have a private queue to do this.
1480 *
1481 * Returns %NULL for an empty list or a pointer to the tail element.
1482 * The reference count is not incremented and the reference is therefore
1483 * volatile. Use with caution.
1484 */
05bdd2f1 1485static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1486{
18d07000
ED
1487 struct sk_buff *skb = list_->prev;
1488
1489 if (skb == (struct sk_buff *)list_)
1490 skb = NULL;
1491 return skb;
1492
1da177e4
LT
1493}
1494
1495/**
1496 * skb_queue_len - get queue length
1497 * @list_: list to measure
1498 *
1499 * Return the length of an &sk_buff queue.
1500 */
1501static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1502{
1503 return list_->qlen;
1504}
1505
67fed459
DM
1506/**
1507 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1508 * @list: queue to initialize
1509 *
1510 * This initializes only the list and queue length aspects of
1511 * an sk_buff_head object. This allows to initialize the list
1512 * aspects of an sk_buff_head without reinitializing things like
1513 * the spinlock. It can also be used for on-stack sk_buff_head
1514 * objects where the spinlock is known to not be used.
1515 */
1516static inline void __skb_queue_head_init(struct sk_buff_head *list)
1517{
1518 list->prev = list->next = (struct sk_buff *)list;
1519 list->qlen = 0;
1520}
1521
76f10ad0
AV
1522/*
1523 * This function creates a split out lock class for each invocation;
1524 * this is needed for now since a whole lot of users of the skb-queue
1525 * infrastructure in drivers have different locking usage (in hardirq)
1526 * than the networking core (in softirq only). In the long run either the
1527 * network layer or drivers should need annotation to consolidate the
1528 * main types of usage into 3 classes.
1529 */
1da177e4
LT
1530static inline void skb_queue_head_init(struct sk_buff_head *list)
1531{
1532 spin_lock_init(&list->lock);
67fed459 1533 __skb_queue_head_init(list);
1da177e4
LT
1534}
1535
c2ecba71
PE
1536static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1537 struct lock_class_key *class)
1538{
1539 skb_queue_head_init(list);
1540 lockdep_set_class(&list->lock, class);
1541}
1542
1da177e4 1543/*
bf299275 1544 * Insert an sk_buff on a list.
1da177e4
LT
1545 *
1546 * The "__skb_xxxx()" functions are the non-atomic ones that
1547 * can only be called with interrupts disabled.
1548 */
7965bd4d
JP
1549void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1550 struct sk_buff_head *list);
bf299275
GR
1551static inline void __skb_insert(struct sk_buff *newsk,
1552 struct sk_buff *prev, struct sk_buff *next,
1553 struct sk_buff_head *list)
1554{
1555 newsk->next = next;
1556 newsk->prev = prev;
1557 next->prev = prev->next = newsk;
1558 list->qlen++;
1559}
1da177e4 1560
67fed459
DM
1561static inline void __skb_queue_splice(const struct sk_buff_head *list,
1562 struct sk_buff *prev,
1563 struct sk_buff *next)
1564{
1565 struct sk_buff *first = list->next;
1566 struct sk_buff *last = list->prev;
1567
1568 first->prev = prev;
1569 prev->next = first;
1570
1571 last->next = next;
1572 next->prev = last;
1573}
1574
1575/**
1576 * skb_queue_splice - join two skb lists, this is designed for stacks
1577 * @list: the new list to add
1578 * @head: the place to add it in the first list
1579 */
1580static inline void skb_queue_splice(const struct sk_buff_head *list,
1581 struct sk_buff_head *head)
1582{
1583 if (!skb_queue_empty(list)) {
1584 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1585 head->qlen += list->qlen;
67fed459
DM
1586 }
1587}
1588
1589/**
d9619496 1590 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1591 * @list: the new list to add
1592 * @head: the place to add it in the first list
1593 *
1594 * The list at @list is reinitialised
1595 */
1596static inline void skb_queue_splice_init(struct sk_buff_head *list,
1597 struct sk_buff_head *head)
1598{
1599 if (!skb_queue_empty(list)) {
1600 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1601 head->qlen += list->qlen;
67fed459
DM
1602 __skb_queue_head_init(list);
1603 }
1604}
1605
1606/**
1607 * skb_queue_splice_tail - join two skb lists, each list being a queue
1608 * @list: the new list to add
1609 * @head: the place to add it in the first list
1610 */
1611static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1612 struct sk_buff_head *head)
1613{
1614 if (!skb_queue_empty(list)) {
1615 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1616 head->qlen += list->qlen;
67fed459
DM
1617 }
1618}
1619
1620/**
d9619496 1621 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1622 * @list: the new list to add
1623 * @head: the place to add it in the first list
1624 *
1625 * Each of the lists is a queue.
1626 * The list at @list is reinitialised
1627 */
1628static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1629 struct sk_buff_head *head)
1630{
1631 if (!skb_queue_empty(list)) {
1632 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1633 head->qlen += list->qlen;
67fed459
DM
1634 __skb_queue_head_init(list);
1635 }
1636}
1637
1da177e4 1638/**
300ce174 1639 * __skb_queue_after - queue a buffer at the list head
1da177e4 1640 * @list: list to use
300ce174 1641 * @prev: place after this buffer
1da177e4
LT
1642 * @newsk: buffer to queue
1643 *
300ce174 1644 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1645 * and you must therefore hold required locks before calling it.
1646 *
1647 * A buffer cannot be placed on two lists at the same time.
1648 */
300ce174
SH
1649static inline void __skb_queue_after(struct sk_buff_head *list,
1650 struct sk_buff *prev,
1651 struct sk_buff *newsk)
1da177e4 1652{
bf299275 1653 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1654}
1655
7965bd4d
JP
1656void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1657 struct sk_buff_head *list);
7de6c033 1658
f5572855
GR
1659static inline void __skb_queue_before(struct sk_buff_head *list,
1660 struct sk_buff *next,
1661 struct sk_buff *newsk)
1662{
1663 __skb_insert(newsk, next->prev, next, list);
1664}
1665
300ce174
SH
1666/**
1667 * __skb_queue_head - queue a buffer at the list head
1668 * @list: list to use
1669 * @newsk: buffer to queue
1670 *
1671 * Queue a buffer at the start of a list. This function takes no locks
1672 * and you must therefore hold required locks before calling it.
1673 *
1674 * A buffer cannot be placed on two lists at the same time.
1675 */
7965bd4d 1676void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174
SH
1677static inline void __skb_queue_head(struct sk_buff_head *list,
1678 struct sk_buff *newsk)
1679{
1680 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1681}
1682
1da177e4
LT
1683/**
1684 * __skb_queue_tail - queue a buffer at the list tail
1685 * @list: list to use
1686 * @newsk: buffer to queue
1687 *
1688 * Queue a buffer at the end of a list. This function takes no locks
1689 * and you must therefore hold required locks before calling it.
1690 *
1691 * A buffer cannot be placed on two lists at the same time.
1692 */
7965bd4d 1693void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4
LT
1694static inline void __skb_queue_tail(struct sk_buff_head *list,
1695 struct sk_buff *newsk)
1696{
f5572855 1697 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4
LT
1698}
1699
1da177e4
LT
1700/*
1701 * remove sk_buff from list. _Must_ be called atomically, and with
1702 * the list known..
1703 */
7965bd4d 1704void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
1705static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1706{
1707 struct sk_buff *next, *prev;
1708
1709 list->qlen--;
1710 next = skb->next;
1711 prev = skb->prev;
1712 skb->next = skb->prev = NULL;
1da177e4
LT
1713 next->prev = prev;
1714 prev->next = next;
1715}
1716
f525c06d
GR
1717/**
1718 * __skb_dequeue - remove from the head of the queue
1719 * @list: list to dequeue from
1720 *
1721 * Remove the head of the list. This function does not take any locks
1722 * so must be used with appropriate locks held only. The head item is
1723 * returned or %NULL if the list is empty.
1724 */
7965bd4d 1725struct sk_buff *skb_dequeue(struct sk_buff_head *list);
f525c06d
GR
1726static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1727{
1728 struct sk_buff *skb = skb_peek(list);
1729 if (skb)
1730 __skb_unlink(skb, list);
1731 return skb;
1732}
1da177e4
LT
1733
1734/**
1735 * __skb_dequeue_tail - remove from the tail of the queue
1736 * @list: list to dequeue from
1737 *
1738 * Remove the tail of the list. This function does not take any locks
1739 * so must be used with appropriate locks held only. The tail item is
1740 * returned or %NULL if the list is empty.
1741 */
7965bd4d 1742struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
1743static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1744{
1745 struct sk_buff *skb = skb_peek_tail(list);
1746 if (skb)
1747 __skb_unlink(skb, list);
1748 return skb;
1749}
1750
1751
bdcc0924 1752static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
1753{
1754 return skb->data_len;
1755}
1756
1757static inline unsigned int skb_headlen(const struct sk_buff *skb)
1758{
1759 return skb->len - skb->data_len;
1760}
1761
c72d8cda 1762static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1da177e4 1763{
c72d8cda 1764 unsigned int i, len = 0;
1da177e4 1765
c72d8cda 1766 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 1767 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
1da177e4
LT
1768 return len + skb_headlen(skb);
1769}
1770
131ea667
IC
1771/**
1772 * __skb_fill_page_desc - initialise a paged fragment in an skb
1773 * @skb: buffer containing fragment to be initialised
1774 * @i: paged fragment index to initialise
1775 * @page: the page to use for this fragment
1776 * @off: the offset to the data with @page
1777 * @size: the length of the data
1778 *
1779 * Initialises the @i'th fragment of @skb to point to &size bytes at
1780 * offset @off within @page.
1781 *
1782 * Does not take any additional reference on the fragment.
1783 */
1784static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1785 struct page *page, int off, int size)
1da177e4
LT
1786{
1787 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1788
c48a11c7 1789 /*
2f064f34
MH
1790 * Propagate page pfmemalloc to the skb if we can. The problem is
1791 * that not all callers have unique ownership of the page but rely
1792 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 1793 */
a8605c60 1794 frag->page.p = page;
1da177e4 1795 frag->page_offset = off;
9e903e08 1796 skb_frag_size_set(frag, size);
cca7af38
PE
1797
1798 page = compound_head(page);
2f064f34 1799 if (page_is_pfmemalloc(page))
cca7af38 1800 skb->pfmemalloc = true;
131ea667
IC
1801}
1802
1803/**
1804 * skb_fill_page_desc - initialise a paged fragment in an skb
1805 * @skb: buffer containing fragment to be initialised
1806 * @i: paged fragment index to initialise
1807 * @page: the page to use for this fragment
1808 * @off: the offset to the data with @page
1809 * @size: the length of the data
1810 *
1811 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 1812 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
1813 * addition updates @skb such that @i is the last fragment.
1814 *
1815 * Does not take any additional reference on the fragment.
1816 */
1817static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1818 struct page *page, int off, int size)
1819{
1820 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
1821 skb_shinfo(skb)->nr_frags = i + 1;
1822}
1823
7965bd4d
JP
1824void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1825 int size, unsigned int truesize);
654bed16 1826
f8e617e1
JW
1827void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1828 unsigned int truesize);
1829
1da177e4 1830#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
21dc3301 1831#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
1da177e4
LT
1832#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1833
27a884dc
ACM
1834#ifdef NET_SKBUFF_DATA_USES_OFFSET
1835static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1836{
1837 return skb->head + skb->tail;
1838}
1839
1840static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1841{
1842 skb->tail = skb->data - skb->head;
1843}
1844
1845static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1846{
1847 skb_reset_tail_pointer(skb);
1848 skb->tail += offset;
1849}
7cc46190 1850
27a884dc
ACM
1851#else /* NET_SKBUFF_DATA_USES_OFFSET */
1852static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1853{
1854 return skb->tail;
1855}
1856
1857static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1858{
1859 skb->tail = skb->data;
1860}
1861
1862static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1863{
1864 skb->tail = skb->data + offset;
1865}
4305b541 1866
27a884dc
ACM
1867#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1868
1da177e4
LT
1869/*
1870 * Add data to an sk_buff
1871 */
0c7ddf36 1872unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
7965bd4d 1873unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1874static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1875{
27a884dc 1876 unsigned char *tmp = skb_tail_pointer(skb);
1da177e4
LT
1877 SKB_LINEAR_ASSERT(skb);
1878 skb->tail += len;
1879 skb->len += len;
1880 return tmp;
1881}
1882
7965bd4d 1883unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1884static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1885{
1886 skb->data -= len;
1887 skb->len += len;
1888 return skb->data;
1889}
1890
7965bd4d 1891unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
1da177e4
LT
1892static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1893{
1894 skb->len -= len;
1895 BUG_ON(skb->len < skb->data_len);
1896 return skb->data += len;
1897}
1898
47d29646
DM
1899static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1900{
1901 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1902}
1903
7965bd4d 1904unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4
LT
1905
1906static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1907{
1908 if (len > skb_headlen(skb) &&
987c402a 1909 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
1910 return NULL;
1911 skb->len -= len;
1912 return skb->data += len;
1913}
1914
1915static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1916{
1917 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1918}
1919
1920static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1921{
1922 if (likely(len <= skb_headlen(skb)))
1923 return 1;
1924 if (unlikely(len > skb->len))
1925 return 0;
987c402a 1926 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
1927}
1928
c8c8b127
ED
1929void skb_condense(struct sk_buff *skb);
1930
1da177e4
LT
1931/**
1932 * skb_headroom - bytes at buffer head
1933 * @skb: buffer to check
1934 *
1935 * Return the number of bytes of free space at the head of an &sk_buff.
1936 */
c2636b4d 1937static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
1938{
1939 return skb->data - skb->head;
1940}
1941
1942/**
1943 * skb_tailroom - bytes at buffer end
1944 * @skb: buffer to check
1945 *
1946 * Return the number of bytes of free space at the tail of an sk_buff
1947 */
1948static inline int skb_tailroom(const struct sk_buff *skb)
1949{
4305b541 1950 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
1951}
1952
a21d4572
ED
1953/**
1954 * skb_availroom - bytes at buffer end
1955 * @skb: buffer to check
1956 *
1957 * Return the number of bytes of free space at the tail of an sk_buff
1958 * allocated by sk_stream_alloc()
1959 */
1960static inline int skb_availroom(const struct sk_buff *skb)
1961{
16fad69c
ED
1962 if (skb_is_nonlinear(skb))
1963 return 0;
1964
1965 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
1966}
1967
1da177e4
LT
1968/**
1969 * skb_reserve - adjust headroom
1970 * @skb: buffer to alter
1971 * @len: bytes to move
1972 *
1973 * Increase the headroom of an empty &sk_buff by reducing the tail
1974 * room. This is only allowed for an empty buffer.
1975 */
8243126c 1976static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
1977{
1978 skb->data += len;
1979 skb->tail += len;
1980}
1981
1837b2e2
BP
1982/**
1983 * skb_tailroom_reserve - adjust reserved_tailroom
1984 * @skb: buffer to alter
1985 * @mtu: maximum amount of headlen permitted
1986 * @needed_tailroom: minimum amount of reserved_tailroom
1987 *
1988 * Set reserved_tailroom so that headlen can be as large as possible but
1989 * not larger than mtu and tailroom cannot be smaller than
1990 * needed_tailroom.
1991 * The required headroom should already have been reserved before using
1992 * this function.
1993 */
1994static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
1995 unsigned int needed_tailroom)
1996{
1997 SKB_LINEAR_ASSERT(skb);
1998 if (mtu < skb_tailroom(skb) - needed_tailroom)
1999 /* use at most mtu */
2000 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2001 else
2002 /* use up to all available space */
2003 skb->reserved_tailroom = needed_tailroom;
2004}
2005
8bce6d7d
TH
2006#define ENCAP_TYPE_ETHER 0
2007#define ENCAP_TYPE_IPPROTO 1
2008
2009static inline void skb_set_inner_protocol(struct sk_buff *skb,
2010 __be16 protocol)
2011{
2012 skb->inner_protocol = protocol;
2013 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2014}
2015
2016static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2017 __u8 ipproto)
2018{
2019 skb->inner_ipproto = ipproto;
2020 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2021}
2022
6a674e9c
JG
2023static inline void skb_reset_inner_headers(struct sk_buff *skb)
2024{
aefbd2b3 2025 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2026 skb->inner_network_header = skb->network_header;
2027 skb->inner_transport_header = skb->transport_header;
2028}
2029
0b5c9db1
JP
2030static inline void skb_reset_mac_len(struct sk_buff *skb)
2031{
2032 skb->mac_len = skb->network_header - skb->mac_header;
2033}
2034
6a674e9c
JG
2035static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2036 *skb)
2037{
2038 return skb->head + skb->inner_transport_header;
2039}
2040
55dc5a9f
TH
2041static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2042{
2043 return skb_inner_transport_header(skb) - skb->data;
2044}
2045
6a674e9c
JG
2046static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2047{
2048 skb->inner_transport_header = skb->data - skb->head;
2049}
2050
2051static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2052 const int offset)
2053{
2054 skb_reset_inner_transport_header(skb);
2055 skb->inner_transport_header += offset;
2056}
2057
2058static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2059{
2060 return skb->head + skb->inner_network_header;
2061}
2062
2063static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2064{
2065 skb->inner_network_header = skb->data - skb->head;
2066}
2067
2068static inline void skb_set_inner_network_header(struct sk_buff *skb,
2069 const int offset)
2070{
2071 skb_reset_inner_network_header(skb);
2072 skb->inner_network_header += offset;
2073}
2074
aefbd2b3
PS
2075static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2076{
2077 return skb->head + skb->inner_mac_header;
2078}
2079
2080static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2081{
2082 skb->inner_mac_header = skb->data - skb->head;
2083}
2084
2085static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2086 const int offset)
2087{
2088 skb_reset_inner_mac_header(skb);
2089 skb->inner_mac_header += offset;
2090}
fda55eca
ED
2091static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2092{
35d04610 2093 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2094}
2095
9c70220b
ACM
2096static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2097{
2e07fa9c 2098 return skb->head + skb->transport_header;
9c70220b
ACM
2099}
2100
badff6d0
ACM
2101static inline void skb_reset_transport_header(struct sk_buff *skb)
2102{
2e07fa9c 2103 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2104}
2105
967b05f6
ACM
2106static inline void skb_set_transport_header(struct sk_buff *skb,
2107 const int offset)
2108{
2e07fa9c
ACM
2109 skb_reset_transport_header(skb);
2110 skb->transport_header += offset;
ea2ae17d
ACM
2111}
2112
d56f90a7
ACM
2113static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2114{
2e07fa9c 2115 return skb->head + skb->network_header;
d56f90a7
ACM
2116}
2117
c1d2bbe1
ACM
2118static inline void skb_reset_network_header(struct sk_buff *skb)
2119{
2e07fa9c 2120 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2121}
2122
c14d2450
ACM
2123static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2124{
2e07fa9c
ACM
2125 skb_reset_network_header(skb);
2126 skb->network_header += offset;
c14d2450
ACM
2127}
2128
2e07fa9c 2129static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2130{
2e07fa9c 2131 return skb->head + skb->mac_header;
bbe735e4
ACM
2132}
2133
ea6da4fd
AV
2134static inline int skb_mac_offset(const struct sk_buff *skb)
2135{
2136 return skb_mac_header(skb) - skb->data;
2137}
2138
2e07fa9c 2139static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2140{
35d04610 2141 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2142}
2143
2144static inline void skb_reset_mac_header(struct sk_buff *skb)
2145{
2146 skb->mac_header = skb->data - skb->head;
2147}
2148
2149static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2150{
2151 skb_reset_mac_header(skb);
2152 skb->mac_header += offset;
2153}
2154
0e3da5bb
TT
2155static inline void skb_pop_mac_header(struct sk_buff *skb)
2156{
2157 skb->mac_header = skb->network_header;
2158}
2159
fbbdb8f0
YX
2160static inline void skb_probe_transport_header(struct sk_buff *skb,
2161 const int offset_hint)
2162{
2163 struct flow_keys keys;
2164
2165 if (skb_transport_header_was_set(skb))
2166 return;
cd79a238 2167 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
42aecaa9 2168 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2169 else
2170 skb_set_transport_header(skb, offset_hint);
2171}
2172
03606895
ED
2173static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2174{
2175 if (skb_mac_header_was_set(skb)) {
2176 const unsigned char *old_mac = skb_mac_header(skb);
2177
2178 skb_set_mac_header(skb, -skb->mac_len);
2179 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2180 }
2181}
2182
04fb451e
MM
2183static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2184{
2185 return skb->csum_start - skb_headroom(skb);
2186}
2187
08b64fcc
AD
2188static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2189{
2190 return skb->head + skb->csum_start;
2191}
2192
2e07fa9c
ACM
2193static inline int skb_transport_offset(const struct sk_buff *skb)
2194{
2195 return skb_transport_header(skb) - skb->data;
2196}
2197
2198static inline u32 skb_network_header_len(const struct sk_buff *skb)
2199{
2200 return skb->transport_header - skb->network_header;
2201}
2202
6a674e9c
JG
2203static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2204{
2205 return skb->inner_transport_header - skb->inner_network_header;
2206}
2207
2e07fa9c
ACM
2208static inline int skb_network_offset(const struct sk_buff *skb)
2209{
2210 return skb_network_header(skb) - skb->data;
2211}
48d49d0c 2212
6a674e9c
JG
2213static inline int skb_inner_network_offset(const struct sk_buff *skb)
2214{
2215 return skb_inner_network_header(skb) - skb->data;
2216}
2217
f9599ce1
CG
2218static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2219{
2220 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2221}
2222
1da177e4
LT
2223/*
2224 * CPUs often take a performance hit when accessing unaligned memory
2225 * locations. The actual performance hit varies, it can be small if the
2226 * hardware handles it or large if we have to take an exception and fix it
2227 * in software.
2228 *
2229 * Since an ethernet header is 14 bytes network drivers often end up with
2230 * the IP header at an unaligned offset. The IP header can be aligned by
2231 * shifting the start of the packet by 2 bytes. Drivers should do this
2232 * with:
2233 *
8660c124 2234 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2235 *
2236 * The downside to this alignment of the IP header is that the DMA is now
2237 * unaligned. On some architectures the cost of an unaligned DMA is high
2238 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2239 *
1da177e4
LT
2240 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2241 * to be overridden.
2242 */
2243#ifndef NET_IP_ALIGN
2244#define NET_IP_ALIGN 2
2245#endif
2246
025be81e
AB
2247/*
2248 * The networking layer reserves some headroom in skb data (via
2249 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2250 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2251 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2252 *
2253 * Unfortunately this headroom changes the DMA alignment of the resulting
2254 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2255 * on some architectures. An architecture can override this value,
2256 * perhaps setting it to a cacheline in size (since that will maintain
2257 * cacheline alignment of the DMA). It must be a power of 2.
2258 *
d6301d3d 2259 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2260 * headroom, you should not reduce this.
5933dd2f
ED
2261 *
2262 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2263 * to reduce average number of cache lines per packet.
2264 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2265 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2266 */
2267#ifndef NET_SKB_PAD
5933dd2f 2268#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2269#endif
2270
7965bd4d 2271int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2272
5293efe6 2273static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2274{
c4264f27 2275 if (unlikely(skb_is_nonlinear(skb))) {
3cc0e873
HX
2276 WARN_ON(1);
2277 return;
2278 }
27a884dc
ACM
2279 skb->len = len;
2280 skb_set_tail_pointer(skb, len);
1da177e4
LT
2281}
2282
5293efe6
DB
2283static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2284{
2285 __skb_set_length(skb, len);
2286}
2287
7965bd4d 2288void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2289
2290static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2291{
3cc0e873
HX
2292 if (skb->data_len)
2293 return ___pskb_trim(skb, len);
2294 __skb_trim(skb, len);
2295 return 0;
1da177e4
LT
2296}
2297
2298static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2299{
2300 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2301}
2302
e9fa4f7b
HX
2303/**
2304 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2305 * @skb: buffer to alter
2306 * @len: new length
2307 *
2308 * This is identical to pskb_trim except that the caller knows that
2309 * the skb is not cloned so we should never get an error due to out-
2310 * of-memory.
2311 */
2312static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2313{
2314 int err = pskb_trim(skb, len);
2315 BUG_ON(err);
2316}
2317
5293efe6
DB
2318static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2319{
2320 unsigned int diff = len - skb->len;
2321
2322 if (skb_tailroom(skb) < diff) {
2323 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2324 GFP_ATOMIC);
2325 if (ret)
2326 return ret;
2327 }
2328 __skb_set_length(skb, len);
2329 return 0;
2330}
2331
1da177e4
LT
2332/**
2333 * skb_orphan - orphan a buffer
2334 * @skb: buffer to orphan
2335 *
2336 * If a buffer currently has an owner then we call the owner's
2337 * destructor function and make the @skb unowned. The buffer continues
2338 * to exist but is no longer charged to its former owner.
2339 */
2340static inline void skb_orphan(struct sk_buff *skb)
2341{
c34a7612 2342 if (skb->destructor) {
1da177e4 2343 skb->destructor(skb);
c34a7612
ED
2344 skb->destructor = NULL;
2345 skb->sk = NULL;
376c7311
ED
2346 } else {
2347 BUG_ON(skb->sk);
c34a7612 2348 }
1da177e4
LT
2349}
2350
a353e0ce
MT
2351/**
2352 * skb_orphan_frags - orphan the frags contained in a buffer
2353 * @skb: buffer to orphan frags from
2354 * @gfp_mask: allocation mask for replacement pages
2355 *
2356 * For each frag in the SKB which needs a destructor (i.e. has an
2357 * owner) create a copy of that frag and release the original
2358 * page by calling the destructor.
2359 */
2360static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2361{
2362 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2363 return 0;
2364 return skb_copy_ubufs(skb, gfp_mask);
2365}
2366
1da177e4
LT
2367/**
2368 * __skb_queue_purge - empty a list
2369 * @list: list to empty
2370 *
2371 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2372 * the list and one reference dropped. This function does not take the
2373 * list lock and the caller must hold the relevant locks to use it.
2374 */
7965bd4d 2375void skb_queue_purge(struct sk_buff_head *list);
1da177e4
LT
2376static inline void __skb_queue_purge(struct sk_buff_head *list)
2377{
2378 struct sk_buff *skb;
2379 while ((skb = __skb_dequeue(list)) != NULL)
2380 kfree_skb(skb);
2381}
2382
9f5afeae
YW
2383void skb_rbtree_purge(struct rb_root *root);
2384
7965bd4d 2385void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2386
7965bd4d
JP
2387struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2388 gfp_t gfp_mask);
8af27456
CH
2389
2390/**
2391 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2392 * @dev: network device to receive on
2393 * @length: length to allocate
2394 *
2395 * Allocate a new &sk_buff and assign it a usage count of one. The
2396 * buffer has unspecified headroom built in. Users should allocate
2397 * the headroom they think they need without accounting for the
2398 * built in space. The built in space is used for optimisations.
2399 *
2400 * %NULL is returned if there is no free memory. Although this function
2401 * allocates memory it can be called from an interrupt.
2402 */
2403static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2404 unsigned int length)
8af27456
CH
2405{
2406 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2407}
2408
6f532612
ED
2409/* legacy helper around __netdev_alloc_skb() */
2410static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2411 gfp_t gfp_mask)
2412{
2413 return __netdev_alloc_skb(NULL, length, gfp_mask);
2414}
2415
2416/* legacy helper around netdev_alloc_skb() */
2417static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2418{
2419 return netdev_alloc_skb(NULL, length);
2420}
2421
2422
4915a0de
ED
2423static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2424 unsigned int length, gfp_t gfp)
61321bbd 2425{
4915a0de 2426 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2427
2428 if (NET_IP_ALIGN && skb)
2429 skb_reserve(skb, NET_IP_ALIGN);
2430 return skb;
2431}
2432
4915a0de
ED
2433static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2434 unsigned int length)
2435{
2436 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2437}
2438
181edb2b
AD
2439static inline void skb_free_frag(void *addr)
2440{
8c2dd3e4 2441 page_frag_free(addr);
181edb2b
AD
2442}
2443
ffde7328 2444void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2445struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2446 unsigned int length, gfp_t gfp_mask);
2447static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2448 unsigned int length)
2449{
2450 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2451}
795bb1c0
JDB
2452void napi_consume_skb(struct sk_buff *skb, int budget);
2453
2454void __kfree_skb_flush(void);
15fad714 2455void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2456
71dfda58
AD
2457/**
2458 * __dev_alloc_pages - allocate page for network Rx
2459 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2460 * @order: size of the allocation
2461 *
2462 * Allocate a new page.
2463 *
2464 * %NULL is returned if there is no free memory.
2465*/
2466static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2467 unsigned int order)
2468{
2469 /* This piece of code contains several assumptions.
2470 * 1. This is for device Rx, therefor a cold page is preferred.
2471 * 2. The expectation is the user wants a compound page.
2472 * 3. If requesting a order 0 page it will not be compound
2473 * due to the check to see if order has a value in prep_new_page
2474 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2475 * code in gfp_to_alloc_flags that should be enforcing this.
2476 */
2477 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2478
2479 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2480}
2481
2482static inline struct page *dev_alloc_pages(unsigned int order)
2483{
95829b3a 2484 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2485}
2486
2487/**
2488 * __dev_alloc_page - allocate a page for network Rx
2489 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2490 *
2491 * Allocate a new page.
2492 *
2493 * %NULL is returned if there is no free memory.
2494 */
2495static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2496{
2497 return __dev_alloc_pages(gfp_mask, 0);
2498}
2499
2500static inline struct page *dev_alloc_page(void)
2501{
95829b3a 2502 return dev_alloc_pages(0);
71dfda58
AD
2503}
2504
0614002b
MG
2505/**
2506 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2507 * @page: The page that was allocated from skb_alloc_page
2508 * @skb: The skb that may need pfmemalloc set
2509 */
2510static inline void skb_propagate_pfmemalloc(struct page *page,
2511 struct sk_buff *skb)
2512{
2f064f34 2513 if (page_is_pfmemalloc(page))
0614002b
MG
2514 skb->pfmemalloc = true;
2515}
2516
131ea667 2517/**
e227867f 2518 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2519 * @frag: the paged fragment
2520 *
2521 * Returns the &struct page associated with @frag.
2522 */
2523static inline struct page *skb_frag_page(const skb_frag_t *frag)
2524{
a8605c60 2525 return frag->page.p;
131ea667
IC
2526}
2527
2528/**
2529 * __skb_frag_ref - take an addition reference on a paged fragment.
2530 * @frag: the paged fragment
2531 *
2532 * Takes an additional reference on the paged fragment @frag.
2533 */
2534static inline void __skb_frag_ref(skb_frag_t *frag)
2535{
2536 get_page(skb_frag_page(frag));
2537}
2538
2539/**
2540 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2541 * @skb: the buffer
2542 * @f: the fragment offset.
2543 *
2544 * Takes an additional reference on the @f'th paged fragment of @skb.
2545 */
2546static inline void skb_frag_ref(struct sk_buff *skb, int f)
2547{
2548 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2549}
2550
2551/**
2552 * __skb_frag_unref - release a reference on a paged fragment.
2553 * @frag: the paged fragment
2554 *
2555 * Releases a reference on the paged fragment @frag.
2556 */
2557static inline void __skb_frag_unref(skb_frag_t *frag)
2558{
2559 put_page(skb_frag_page(frag));
2560}
2561
2562/**
2563 * skb_frag_unref - release a reference on a paged fragment of an skb.
2564 * @skb: the buffer
2565 * @f: the fragment offset
2566 *
2567 * Releases a reference on the @f'th paged fragment of @skb.
2568 */
2569static inline void skb_frag_unref(struct sk_buff *skb, int f)
2570{
2571 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2572}
2573
2574/**
2575 * skb_frag_address - gets the address of the data contained in a paged fragment
2576 * @frag: the paged fragment buffer
2577 *
2578 * Returns the address of the data within @frag. The page must already
2579 * be mapped.
2580 */
2581static inline void *skb_frag_address(const skb_frag_t *frag)
2582{
2583 return page_address(skb_frag_page(frag)) + frag->page_offset;
2584}
2585
2586/**
2587 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2588 * @frag: the paged fragment buffer
2589 *
2590 * Returns the address of the data within @frag. Checks that the page
2591 * is mapped and returns %NULL otherwise.
2592 */
2593static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2594{
2595 void *ptr = page_address(skb_frag_page(frag));
2596 if (unlikely(!ptr))
2597 return NULL;
2598
2599 return ptr + frag->page_offset;
2600}
2601
2602/**
2603 * __skb_frag_set_page - sets the page contained in a paged fragment
2604 * @frag: the paged fragment
2605 * @page: the page to set
2606 *
2607 * Sets the fragment @frag to contain @page.
2608 */
2609static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2610{
a8605c60 2611 frag->page.p = page;
131ea667
IC
2612}
2613
2614/**
2615 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2616 * @skb: the buffer
2617 * @f: the fragment offset
2618 * @page: the page to set
2619 *
2620 * Sets the @f'th fragment of @skb to contain @page.
2621 */
2622static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2623 struct page *page)
2624{
2625 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2626}
2627
400dfd3a
ED
2628bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2629
131ea667
IC
2630/**
2631 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 2632 * @dev: the device to map the fragment to
131ea667
IC
2633 * @frag: the paged fragment to map
2634 * @offset: the offset within the fragment (starting at the
2635 * fragment's own offset)
2636 * @size: the number of bytes to map
f83347df 2637 * @dir: the direction of the mapping (%PCI_DMA_*)
131ea667
IC
2638 *
2639 * Maps the page associated with @frag to @device.
2640 */
2641static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2642 const skb_frag_t *frag,
2643 size_t offset, size_t size,
2644 enum dma_data_direction dir)
2645{
2646 return dma_map_page(dev, skb_frag_page(frag),
2647 frag->page_offset + offset, size, dir);
2648}
2649
117632e6
ED
2650static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2651 gfp_t gfp_mask)
2652{
2653 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2654}
2655
bad93e9d
OP
2656
2657static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2658 gfp_t gfp_mask)
2659{
2660 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2661}
2662
2663
334a8132
PM
2664/**
2665 * skb_clone_writable - is the header of a clone writable
2666 * @skb: buffer to check
2667 * @len: length up to which to write
2668 *
2669 * Returns true if modifying the header part of the cloned buffer
2670 * does not requires the data to be copied.
2671 */
05bdd2f1 2672static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
2673{
2674 return !skb_header_cloned(skb) &&
2675 skb_headroom(skb) + len <= skb->hdr_len;
2676}
2677
3697649f
DB
2678static inline int skb_try_make_writable(struct sk_buff *skb,
2679 unsigned int write_len)
2680{
2681 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2682 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2683}
2684
d9cc2048
HX
2685static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2686 int cloned)
2687{
2688 int delta = 0;
2689
d9cc2048
HX
2690 if (headroom > skb_headroom(skb))
2691 delta = headroom - skb_headroom(skb);
2692
2693 if (delta || cloned)
2694 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2695 GFP_ATOMIC);
2696 return 0;
2697}
2698
1da177e4
LT
2699/**
2700 * skb_cow - copy header of skb when it is required
2701 * @skb: buffer to cow
2702 * @headroom: needed headroom
2703 *
2704 * If the skb passed lacks sufficient headroom or its data part
2705 * is shared, data is reallocated. If reallocation fails, an error
2706 * is returned and original skb is not changed.
2707 *
2708 * The result is skb with writable area skb->head...skb->tail
2709 * and at least @headroom of space at head.
2710 */
2711static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2712{
d9cc2048
HX
2713 return __skb_cow(skb, headroom, skb_cloned(skb));
2714}
1da177e4 2715
d9cc2048
HX
2716/**
2717 * skb_cow_head - skb_cow but only making the head writable
2718 * @skb: buffer to cow
2719 * @headroom: needed headroom
2720 *
2721 * This function is identical to skb_cow except that we replace the
2722 * skb_cloned check by skb_header_cloned. It should be used when
2723 * you only need to push on some header and do not need to modify
2724 * the data.
2725 */
2726static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2727{
2728 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
2729}
2730
2731/**
2732 * skb_padto - pad an skbuff up to a minimal size
2733 * @skb: buffer to pad
2734 * @len: minimal length
2735 *
2736 * Pads up a buffer to ensure the trailing bytes exist and are
2737 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
2738 * is untouched. Otherwise it is extended. Returns zero on
2739 * success. The skb is freed on error.
1da177e4 2740 */
5b057c6b 2741static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2742{
2743 unsigned int size = skb->len;
2744 if (likely(size >= len))
5b057c6b 2745 return 0;
987c402a 2746 return skb_pad(skb, len - size);
1da177e4
LT
2747}
2748
9c0c1124
AD
2749/**
2750 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2751 * @skb: buffer to pad
2752 * @len: minimal length
2753 *
2754 * Pads up a buffer to ensure the trailing bytes exist and are
2755 * blanked. If the buffer already contains sufficient data it
2756 * is untouched. Otherwise it is extended. Returns zero on
2757 * success. The skb is freed on error.
2758 */
2759static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2760{
2761 unsigned int size = skb->len;
2762
2763 if (unlikely(size < len)) {
2764 len -= size;
2765 if (skb_pad(skb, len))
2766 return -ENOMEM;
2767 __skb_put(skb, len);
2768 }
2769 return 0;
2770}
2771
1da177e4 2772static inline int skb_add_data(struct sk_buff *skb,
af2b040e 2773 struct iov_iter *from, int copy)
1da177e4
LT
2774{
2775 const int off = skb->len;
2776
2777 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 2778 __wsum csum = 0;
15e6cb46
AV
2779 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2780 &csum, from)) {
1da177e4
LT
2781 skb->csum = csum_block_add(skb->csum, csum, off);
2782 return 0;
2783 }
15e6cb46 2784 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
2785 return 0;
2786
2787 __skb_trim(skb, off);
2788 return -EFAULT;
2789}
2790
38ba0a65
ED
2791static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2792 const struct page *page, int off)
1da177e4
LT
2793{
2794 if (i) {
9e903e08 2795 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 2796
ea2ab693 2797 return page == skb_frag_page(frag) &&
9e903e08 2798 off == frag->page_offset + skb_frag_size(frag);
1da177e4 2799 }
38ba0a65 2800 return false;
1da177e4
LT
2801}
2802
364c6bad
HX
2803static inline int __skb_linearize(struct sk_buff *skb)
2804{
2805 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2806}
2807
1da177e4
LT
2808/**
2809 * skb_linearize - convert paged skb to linear one
2810 * @skb: buffer to linarize
1da177e4
LT
2811 *
2812 * If there is no free memory -ENOMEM is returned, otherwise zero
2813 * is returned and the old skb data released.
2814 */
364c6bad
HX
2815static inline int skb_linearize(struct sk_buff *skb)
2816{
2817 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2818}
2819
cef401de
ED
2820/**
2821 * skb_has_shared_frag - can any frag be overwritten
2822 * @skb: buffer to test
2823 *
2824 * Return true if the skb has at least one frag that might be modified
2825 * by an external entity (as in vmsplice()/sendfile())
2826 */
2827static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2828{
c9af6db4
PS
2829 return skb_is_nonlinear(skb) &&
2830 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
2831}
2832
364c6bad
HX
2833/**
2834 * skb_linearize_cow - make sure skb is linear and writable
2835 * @skb: buffer to process
2836 *
2837 * If there is no free memory -ENOMEM is returned, otherwise zero
2838 * is returned and the old skb data released.
2839 */
2840static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 2841{
364c6bad
HX
2842 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2843 __skb_linearize(skb) : 0;
1da177e4
LT
2844}
2845
479ffccc
DB
2846static __always_inline void
2847__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2848 unsigned int off)
2849{
2850 if (skb->ip_summed == CHECKSUM_COMPLETE)
2851 skb->csum = csum_block_sub(skb->csum,
2852 csum_partial(start, len, 0), off);
2853 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2854 skb_checksum_start_offset(skb) < 0)
2855 skb->ip_summed = CHECKSUM_NONE;
2856}
2857
1da177e4
LT
2858/**
2859 * skb_postpull_rcsum - update checksum for received skb after pull
2860 * @skb: buffer to update
2861 * @start: start of data before pull
2862 * @len: length of data pulled
2863 *
2864 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
2865 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2866 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 2867 */
1da177e4 2868static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 2869 const void *start, unsigned int len)
1da177e4 2870{
479ffccc 2871 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
2872}
2873
479ffccc
DB
2874static __always_inline void
2875__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2876 unsigned int off)
2877{
2878 if (skb->ip_summed == CHECKSUM_COMPLETE)
2879 skb->csum = csum_block_add(skb->csum,
2880 csum_partial(start, len, 0), off);
2881}
cbb042f9 2882
479ffccc
DB
2883/**
2884 * skb_postpush_rcsum - update checksum for received skb after push
2885 * @skb: buffer to update
2886 * @start: start of data after push
2887 * @len: length of data pushed
2888 *
2889 * After doing a push on a received packet, you need to call this to
2890 * update the CHECKSUM_COMPLETE checksum.
2891 */
f8ffad69
DB
2892static inline void skb_postpush_rcsum(struct sk_buff *skb,
2893 const void *start, unsigned int len)
2894{
479ffccc 2895 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
2896}
2897
479ffccc
DB
2898unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2899
82a31b92
WC
2900/**
2901 * skb_push_rcsum - push skb and update receive checksum
2902 * @skb: buffer to update
2903 * @len: length of data pulled
2904 *
2905 * This function performs an skb_push on the packet and updates
2906 * the CHECKSUM_COMPLETE checksum. It should be used on
2907 * receive path processing instead of skb_push unless you know
2908 * that the checksum difference is zero (e.g., a valid IP header)
2909 * or you are setting ip_summed to CHECKSUM_NONE.
2910 */
2911static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2912 unsigned int len)
2913{
2914 skb_push(skb, len);
2915 skb_postpush_rcsum(skb, skb->data, len);
2916 return skb->data;
2917}
2918
7ce5a27f
DM
2919/**
2920 * pskb_trim_rcsum - trim received skb and update checksum
2921 * @skb: buffer to trim
2922 * @len: new length
2923 *
2924 * This is exactly the same as pskb_trim except that it ensures the
2925 * checksum of received packets are still valid after the operation.
2926 */
2927
2928static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2929{
2930 if (likely(len >= skb->len))
2931 return 0;
2932 if (skb->ip_summed == CHECKSUM_COMPLETE)
2933 skb->ip_summed = CHECKSUM_NONE;
2934 return __pskb_trim(skb, len);
2935}
2936
5293efe6
DB
2937static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2938{
2939 if (skb->ip_summed == CHECKSUM_COMPLETE)
2940 skb->ip_summed = CHECKSUM_NONE;
2941 __skb_trim(skb, len);
2942 return 0;
2943}
2944
2945static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2946{
2947 if (skb->ip_summed == CHECKSUM_COMPLETE)
2948 skb->ip_summed = CHECKSUM_NONE;
2949 return __skb_grow(skb, len);
2950}
2951
1da177e4
LT
2952#define skb_queue_walk(queue, skb) \
2953 for (skb = (queue)->next; \
a1e4891f 2954 skb != (struct sk_buff *)(queue); \
1da177e4
LT
2955 skb = skb->next)
2956
46f8914e
JC
2957#define skb_queue_walk_safe(queue, skb, tmp) \
2958 for (skb = (queue)->next, tmp = skb->next; \
2959 skb != (struct sk_buff *)(queue); \
2960 skb = tmp, tmp = skb->next)
2961
1164f52a 2962#define skb_queue_walk_from(queue, skb) \
a1e4891f 2963 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
2964 skb = skb->next)
2965
2966#define skb_queue_walk_from_safe(queue, skb, tmp) \
2967 for (tmp = skb->next; \
2968 skb != (struct sk_buff *)(queue); \
2969 skb = tmp, tmp = skb->next)
2970
300ce174
SH
2971#define skb_queue_reverse_walk(queue, skb) \
2972 for (skb = (queue)->prev; \
a1e4891f 2973 skb != (struct sk_buff *)(queue); \
300ce174
SH
2974 skb = skb->prev)
2975
686a2955
DM
2976#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2977 for (skb = (queue)->prev, tmp = skb->prev; \
2978 skb != (struct sk_buff *)(queue); \
2979 skb = tmp, tmp = skb->prev)
2980
2981#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2982 for (tmp = skb->prev; \
2983 skb != (struct sk_buff *)(queue); \
2984 skb = tmp, tmp = skb->prev)
1da177e4 2985
21dc3301 2986static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
2987{
2988 return skb_shinfo(skb)->frag_list != NULL;
2989}
2990
2991static inline void skb_frag_list_init(struct sk_buff *skb)
2992{
2993 skb_shinfo(skb)->frag_list = NULL;
2994}
2995
ee039871
DM
2996#define skb_walk_frags(skb, iter) \
2997 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2998
ea3793ee
RW
2999
3000int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3001 const struct sk_buff *skb);
65101aec
PA
3002struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3003 struct sk_buff_head *queue,
3004 unsigned int flags,
3005 void (*destructor)(struct sock *sk,
3006 struct sk_buff *skb),
3007 int *peeked, int *off, int *err,
3008 struct sk_buff **last);
ea3793ee 3009struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3010 void (*destructor)(struct sock *sk,
3011 struct sk_buff *skb),
ea3793ee
RW
3012 int *peeked, int *off, int *err,
3013 struct sk_buff **last);
7965bd4d 3014struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3015 void (*destructor)(struct sock *sk,
3016 struct sk_buff *skb),
7965bd4d
JP
3017 int *peeked, int *off, int *err);
3018struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3019 int *err);
3020unsigned int datagram_poll(struct file *file, struct socket *sock,
3021 struct poll_table_struct *wait);
c0371da6
AV
3022int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3023 struct iov_iter *to, int size);
51f3d02b
DM
3024static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3025 struct msghdr *msg, int size)
3026{
e5a4b0bb 3027 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3028}
e5a4b0bb
AV
3029int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3030 struct msghdr *msg);
3a654f97
AV
3031int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3032 struct iov_iter *from, int len);
3a654f97 3033int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3034void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3035void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3036static inline void skb_free_datagram_locked(struct sock *sk,
3037 struct sk_buff *skb)
3038{
3039 __skb_free_datagram_locked(sk, skb, 0);
3040}
7965bd4d 3041int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3042int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3043int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3044__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3045 int len, __wsum csum);
a60e3cc7 3046int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3047 struct pipe_inode_info *pipe, unsigned int len,
25869262 3048 unsigned int flags);
7965bd4d 3049void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3050unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3051int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3052 int len, int hlen);
7965bd4d
JP
3053void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3054int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3055void skb_scrub_packet(struct sk_buff *skb, bool xnet);
de960aa9 3056unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
ae7ef81e 3057bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
7965bd4d 3058struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3059struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3060int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3061int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3062int skb_vlan_pop(struct sk_buff *skb);
3063int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
6fa01ccd
SV
3064struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3065 gfp_t gfp);
20380731 3066
6ce8e9ce
AV
3067static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3068{
3073f070 3069 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3070}
3071
7eab8d9e
AV
3072static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3073{
e5a4b0bb 3074 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3075}
3076
2817a336
DB
3077struct skb_checksum_ops {
3078 __wsum (*update)(const void *mem, int len, __wsum wsum);
3079 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3080};
3081
9617813d
DC
3082extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3083
2817a336
DB
3084__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3085 __wsum csum, const struct skb_checksum_ops *ops);
3086__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3087 __wsum csum);
3088
1e98a0f0
ED
3089static inline void * __must_check
3090__skb_header_pointer(const struct sk_buff *skb, int offset,
3091 int len, void *data, int hlen, void *buffer)
1da177e4 3092{
55820ee2 3093 if (hlen - offset >= len)
690e36e7 3094 return data + offset;
1da177e4 3095
690e36e7
DM
3096 if (!skb ||
3097 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3098 return NULL;
3099
3100 return buffer;
3101}
3102
1e98a0f0
ED
3103static inline void * __must_check
3104skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3105{
3106 return __skb_header_pointer(skb, offset, len, skb->data,
3107 skb_headlen(skb), buffer);
3108}
3109
4262e5cc
DB
3110/**
3111 * skb_needs_linearize - check if we need to linearize a given skb
3112 * depending on the given device features.
3113 * @skb: socket buffer to check
3114 * @features: net device features
3115 *
3116 * Returns true if either:
3117 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3118 * 2. skb is fragmented and the device does not support SG.
3119 */
3120static inline bool skb_needs_linearize(struct sk_buff *skb,
3121 netdev_features_t features)
3122{
3123 return skb_is_nonlinear(skb) &&
3124 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3125 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3126}
3127
d626f62b
ACM
3128static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3129 void *to,
3130 const unsigned int len)
3131{
3132 memcpy(to, skb->data, len);
3133}
3134
3135static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3136 const int offset, void *to,
3137 const unsigned int len)
3138{
3139 memcpy(to, skb->data + offset, len);
3140}
3141
27d7ff46
ACM
3142static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3143 const void *from,
3144 const unsigned int len)
3145{
3146 memcpy(skb->data, from, len);
3147}
3148
3149static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3150 const int offset,
3151 const void *from,
3152 const unsigned int len)
3153{
3154 memcpy(skb->data + offset, from, len);
3155}
3156
7965bd4d 3157void skb_init(void);
1da177e4 3158
ac45f602
PO
3159static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3160{
3161 return skb->tstamp;
3162}
3163
a61bbcf2
PM
3164/**
3165 * skb_get_timestamp - get timestamp from a skb
3166 * @skb: skb to get stamp from
3167 * @stamp: pointer to struct timeval to store stamp in
3168 *
3169 * Timestamps are stored in the skb as offsets to a base timestamp.
3170 * This function converts the offset back to a struct timeval and stores
3171 * it in stamp.
3172 */
ac45f602
PO
3173static inline void skb_get_timestamp(const struct sk_buff *skb,
3174 struct timeval *stamp)
a61bbcf2 3175{
b7aa0bf7 3176 *stamp = ktime_to_timeval(skb->tstamp);
a61bbcf2
PM
3177}
3178
ac45f602
PO
3179static inline void skb_get_timestampns(const struct sk_buff *skb,
3180 struct timespec *stamp)
3181{
3182 *stamp = ktime_to_timespec(skb->tstamp);
3183}
3184
b7aa0bf7 3185static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3186{
b7aa0bf7 3187 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3188}
3189
164891aa
SH
3190static inline ktime_t net_timedelta(ktime_t t)
3191{
3192 return ktime_sub(ktime_get_real(), t);
3193}
3194
b9ce204f
IJ
3195static inline ktime_t net_invalid_timestamp(void)
3196{
8b0e1953 3197 return 0;
b9ce204f 3198}
a61bbcf2 3199
62bccb8c
AD
3200struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3201
c1f19b51
RC
3202#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3203
7965bd4d
JP
3204void skb_clone_tx_timestamp(struct sk_buff *skb);
3205bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3206
3207#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3208
3209static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3210{
3211}
3212
3213static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3214{
3215 return false;
3216}
3217
3218#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3219
3220/**
3221 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3222 *
da92b194
RC
3223 * PHY drivers may accept clones of transmitted packets for
3224 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3225 * must call this function to return the skb back to the stack with a
3226 * timestamp.
da92b194 3227 *
c1f19b51 3228 * @skb: clone of the the original outgoing packet
7a76a021 3229 * @hwtstamps: hardware time stamps
c1f19b51
RC
3230 *
3231 */
3232void skb_complete_tx_timestamp(struct sk_buff *skb,
3233 struct skb_shared_hwtstamps *hwtstamps);
3234
e7fd2885
WB
3235void __skb_tstamp_tx(struct sk_buff *orig_skb,
3236 struct skb_shared_hwtstamps *hwtstamps,
3237 struct sock *sk, int tstype);
3238
ac45f602
PO
3239/**
3240 * skb_tstamp_tx - queue clone of skb with send time stamps
3241 * @orig_skb: the original outgoing packet
3242 * @hwtstamps: hardware time stamps, may be NULL if not available
3243 *
3244 * If the skb has a socket associated, then this function clones the
3245 * skb (thus sharing the actual data and optional structures), stores
3246 * the optional hardware time stamping information (if non NULL) or
3247 * generates a software time stamp (otherwise), then queues the clone
3248 * to the error queue of the socket. Errors are silently ignored.
3249 */
7965bd4d
JP
3250void skb_tstamp_tx(struct sk_buff *orig_skb,
3251 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3252
4507a715
RC
3253static inline void sw_tx_timestamp(struct sk_buff *skb)
3254{
2244d07b
OH
3255 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3256 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
4507a715
RC
3257 skb_tstamp_tx(skb, NULL);
3258}
3259
3260/**
3261 * skb_tx_timestamp() - Driver hook for transmit timestamping
3262 *
3263 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3264 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3265 *
73409f3b
DM
3266 * Specifically, one should make absolutely sure that this function is
3267 * called before TX completion of this packet can trigger. Otherwise
3268 * the packet could potentially already be freed.
3269 *
4507a715
RC
3270 * @skb: A socket buffer.
3271 */
3272static inline void skb_tx_timestamp(struct sk_buff *skb)
3273{
c1f19b51 3274 skb_clone_tx_timestamp(skb);
4507a715
RC
3275 sw_tx_timestamp(skb);
3276}
3277
6e3e939f
JB
3278/**
3279 * skb_complete_wifi_ack - deliver skb with wifi status
3280 *
3281 * @skb: the original outgoing packet
3282 * @acked: ack status
3283 *
3284 */
3285void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3286
7965bd4d
JP
3287__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3288__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3289
60476372
HX
3290static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3291{
6edec0e6
TH
3292 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3293 skb->csum_valid ||
3294 (skb->ip_summed == CHECKSUM_PARTIAL &&
3295 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3296}
3297
fb286bb2
HX
3298/**
3299 * skb_checksum_complete - Calculate checksum of an entire packet
3300 * @skb: packet to process
3301 *
3302 * This function calculates the checksum over the entire packet plus
3303 * the value of skb->csum. The latter can be used to supply the
3304 * checksum of a pseudo header as used by TCP/UDP. It returns the
3305 * checksum.
3306 *
3307 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3308 * this function can be used to verify that checksum on received
3309 * packets. In that case the function should return zero if the
3310 * checksum is correct. In particular, this function will return zero
3311 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3312 * hardware has already verified the correctness of the checksum.
3313 */
4381ca3c 3314static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3315{
60476372
HX
3316 return skb_csum_unnecessary(skb) ?
3317 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3318}
3319
77cffe23
TH
3320static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3321{
3322 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3323 if (skb->csum_level == 0)
3324 skb->ip_summed = CHECKSUM_NONE;
3325 else
3326 skb->csum_level--;
3327 }
3328}
3329
3330static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3331{
3332 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3333 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3334 skb->csum_level++;
3335 } else if (skb->ip_summed == CHECKSUM_NONE) {
3336 skb->ip_summed = CHECKSUM_UNNECESSARY;
3337 skb->csum_level = 0;
3338 }
3339}
3340
76ba0aae
TH
3341/* Check if we need to perform checksum complete validation.
3342 *
3343 * Returns true if checksum complete is needed, false otherwise
3344 * (either checksum is unnecessary or zero checksum is allowed).
3345 */
3346static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3347 bool zero_okay,
3348 __sum16 check)
3349{
5d0c2b95
TH
3350 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3351 skb->csum_valid = 1;
77cffe23 3352 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3353 return false;
3354 }
3355
3356 return true;
3357}
3358
3359/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3360 * in checksum_init.
3361 */
3362#define CHECKSUM_BREAK 76
3363
4e18b9ad
TH
3364/* Unset checksum-complete
3365 *
3366 * Unset checksum complete can be done when packet is being modified
3367 * (uncompressed for instance) and checksum-complete value is
3368 * invalidated.
3369 */
3370static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3371{
3372 if (skb->ip_summed == CHECKSUM_COMPLETE)
3373 skb->ip_summed = CHECKSUM_NONE;
3374}
3375
76ba0aae
TH
3376/* Validate (init) checksum based on checksum complete.
3377 *
3378 * Return values:
3379 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3380 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3381 * checksum is stored in skb->csum for use in __skb_checksum_complete
3382 * non-zero: value of invalid checksum
3383 *
3384 */
3385static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3386 bool complete,
3387 __wsum psum)
3388{
3389 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3390 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3391 skb->csum_valid = 1;
76ba0aae
TH
3392 return 0;
3393 }
3394 }
3395
3396 skb->csum = psum;
3397
5d0c2b95
TH
3398 if (complete || skb->len <= CHECKSUM_BREAK) {
3399 __sum16 csum;
3400
3401 csum = __skb_checksum_complete(skb);
3402 skb->csum_valid = !csum;
3403 return csum;
3404 }
76ba0aae
TH
3405
3406 return 0;
3407}
3408
3409static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3410{
3411 return 0;
3412}
3413
3414/* Perform checksum validate (init). Note that this is a macro since we only
3415 * want to calculate the pseudo header which is an input function if necessary.
3416 * First we try to validate without any computation (checksum unnecessary) and
3417 * then calculate based on checksum complete calling the function to compute
3418 * pseudo header.
3419 *
3420 * Return values:
3421 * 0: checksum is validated or try to in skb_checksum_complete
3422 * non-zero: value of invalid checksum
3423 */
3424#define __skb_checksum_validate(skb, proto, complete, \
3425 zero_okay, check, compute_pseudo) \
3426({ \
3427 __sum16 __ret = 0; \
5d0c2b95 3428 skb->csum_valid = 0; \
76ba0aae
TH
3429 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3430 __ret = __skb_checksum_validate_complete(skb, \
3431 complete, compute_pseudo(skb, proto)); \
3432 __ret; \
3433})
3434
3435#define skb_checksum_init(skb, proto, compute_pseudo) \
3436 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3437
3438#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3439 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3440
3441#define skb_checksum_validate(skb, proto, compute_pseudo) \
3442 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3443
3444#define skb_checksum_validate_zero_check(skb, proto, check, \
3445 compute_pseudo) \
096a4cfa 3446 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
3447
3448#define skb_checksum_simple_validate(skb) \
3449 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3450
d96535a1
TH
3451static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3452{
219f1d79 3453 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
3454}
3455
3456static inline void __skb_checksum_convert(struct sk_buff *skb,
3457 __sum16 check, __wsum pseudo)
3458{
3459 skb->csum = ~pseudo;
3460 skb->ip_summed = CHECKSUM_COMPLETE;
3461}
3462
3463#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3464do { \
3465 if (__skb_checksum_convert_check(skb)) \
3466 __skb_checksum_convert(skb, check, \
3467 compute_pseudo(skb, proto)); \
3468} while (0)
3469
15e2396d
TH
3470static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3471 u16 start, u16 offset)
3472{
3473 skb->ip_summed = CHECKSUM_PARTIAL;
3474 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3475 skb->csum_offset = offset - start;
3476}
3477
dcdc8994
TH
3478/* Update skbuf and packet to reflect the remote checksum offload operation.
3479 * When called, ptr indicates the starting point for skb->csum when
3480 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3481 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3482 */
3483static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 3484 int start, int offset, bool nopartial)
dcdc8994
TH
3485{
3486 __wsum delta;
3487
15e2396d
TH
3488 if (!nopartial) {
3489 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3490 return;
3491 }
3492
dcdc8994
TH
3493 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3494 __skb_checksum_complete(skb);
3495 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3496 }
3497
3498 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3499
3500 /* Adjust skb->csum since we changed the packet */
3501 skb->csum = csum_add(skb->csum, delta);
3502}
3503
cb9c6836
FW
3504static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3505{
3506#if IS_ENABLED(CONFIG_NF_CONNTRACK)
a9e419dc 3507 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
cb9c6836
FW
3508#else
3509 return NULL;
3510#endif
3511}
3512
5f79e0f9 3513#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
7965bd4d 3514void nf_conntrack_destroy(struct nf_conntrack *nfct);
1da177e4
LT
3515static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3516{
3517 if (nfct && atomic_dec_and_test(&nfct->use))
de6e05c4 3518 nf_conntrack_destroy(nfct);
1da177e4
LT
3519}
3520static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3521{
3522 if (nfct)
3523 atomic_inc(&nfct->use);
3524}
2fc72c7b 3525#endif
34666d46 3526#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4
LT
3527static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3528{
3529 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3530 kfree(nf_bridge);
3531}
3532static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3533{
3534 if (nf_bridge)
3535 atomic_inc(&nf_bridge->use);
3536}
3537#endif /* CONFIG_BRIDGE_NETFILTER */
a193a4ab
PM
3538static inline void nf_reset(struct sk_buff *skb)
3539{
5f79e0f9 3540#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3541 nf_conntrack_put(skb_nfct(skb));
3542 skb->_nfct = 0;
2fc72c7b 3543#endif
34666d46 3544#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
a193a4ab
PM
3545 nf_bridge_put(skb->nf_bridge);
3546 skb->nf_bridge = NULL;
3547#endif
3548}
3549
124dff01
PM
3550static inline void nf_reset_trace(struct sk_buff *skb)
3551{
478b360a 3552#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
3553 skb->nf_trace = 0;
3554#endif
a193a4ab
PM
3555}
3556
edda553c 3557/* Note: This doesn't put any conntrack and bridge info in dst. */
b1937227
ED
3558static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3559 bool copy)
edda553c 3560{
5f79e0f9 3561#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
3562 dst->_nfct = src->_nfct;
3563 nf_conntrack_get(skb_nfct(src));
2fc72c7b 3564#endif
34666d46 3565#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
edda553c
YK
3566 dst->nf_bridge = src->nf_bridge;
3567 nf_bridge_get(src->nf_bridge);
3568#endif
478b360a 3569#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
3570 if (copy)
3571 dst->nf_trace = src->nf_trace;
478b360a 3572#endif
edda553c
YK
3573}
3574
e7ac05f3
YK
3575static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3576{
e7ac05f3 3577#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 3578 nf_conntrack_put(skb_nfct(dst));
2fc72c7b 3579#endif
34666d46 3580#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
e7ac05f3
YK
3581 nf_bridge_put(dst->nf_bridge);
3582#endif
b1937227 3583 __nf_copy(dst, src, true);
e7ac05f3
YK
3584}
3585
984bc16c
JM
3586#ifdef CONFIG_NETWORK_SECMARK
3587static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3588{
3589 to->secmark = from->secmark;
3590}
3591
3592static inline void skb_init_secmark(struct sk_buff *skb)
3593{
3594 skb->secmark = 0;
3595}
3596#else
3597static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3598{ }
3599
3600static inline void skb_init_secmark(struct sk_buff *skb)
3601{ }
3602#endif
3603
574f7194
EB
3604static inline bool skb_irq_freeable(const struct sk_buff *skb)
3605{
3606 return !skb->destructor &&
3607#if IS_ENABLED(CONFIG_XFRM)
3608 !skb->sp &&
3609#endif
cb9c6836 3610 !skb_nfct(skb) &&
574f7194
EB
3611 !skb->_skb_refdst &&
3612 !skb_has_frag_list(skb);
3613}
3614
f25f4e44
PWJ
3615static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3616{
f25f4e44 3617 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
3618}
3619
9247744e 3620static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 3621{
4e3ab47a 3622 return skb->queue_mapping;
4e3ab47a
PE
3623}
3624
f25f4e44
PWJ
3625static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3626{
f25f4e44 3627 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
3628}
3629
d5a9e24a
DM
3630static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3631{
3632 skb->queue_mapping = rx_queue + 1;
3633}
3634
9247744e 3635static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
3636{
3637 return skb->queue_mapping - 1;
3638}
3639
9247744e 3640static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 3641{
a02cec21 3642 return skb->queue_mapping != 0;
d5a9e24a
DM
3643}
3644
4ff06203
JA
3645static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3646{
3647 skb->dst_pending_confirm = val;
3648}
3649
3650static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3651{
3652 return skb->dst_pending_confirm != 0;
3653}
3654
def8b4fa
AD
3655static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3656{
0b3d8e08 3657#ifdef CONFIG_XFRM
def8b4fa 3658 return skb->sp;
def8b4fa 3659#else
def8b4fa 3660 return NULL;
def8b4fa 3661#endif
0b3d8e08 3662}
def8b4fa 3663
68c33163
PS
3664/* Keeps track of mac header offset relative to skb->head.
3665 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3666 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
3667 * tunnel skb it points to outer mac header.
3668 * Keeps track of level of encapsulation of network headers.
3669 */
68c33163 3670struct skb_gso_cb {
802ab55a
AD
3671 union {
3672 int mac_offset;
3673 int data_offset;
3674 };
3347c960 3675 int encap_level;
76443456 3676 __wsum csum;
7e2b10c1 3677 __u16 csum_start;
68c33163 3678};
9207f9d4
KK
3679#define SKB_SGO_CB_OFFSET 32
3680#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
3681
3682static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3683{
3684 return (skb_mac_header(inner_skb) - inner_skb->head) -
3685 SKB_GSO_CB(inner_skb)->mac_offset;
3686}
3687
1e2bd517
PS
3688static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3689{
3690 int new_headroom, headroom;
3691 int ret;
3692
3693 headroom = skb_headroom(skb);
3694 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3695 if (ret)
3696 return ret;
3697
3698 new_headroom = skb_headroom(skb);
3699 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3700 return 0;
3701}
3702
08b64fcc
AD
3703static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3704{
3705 /* Do not update partial checksums if remote checksum is enabled. */
3706 if (skb->remcsum_offload)
3707 return;
3708
3709 SKB_GSO_CB(skb)->csum = res;
3710 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3711}
3712
7e2b10c1
TH
3713/* Compute the checksum for a gso segment. First compute the checksum value
3714 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3715 * then add in skb->csum (checksum from csum_start to end of packet).
3716 * skb->csum and csum_start are then updated to reflect the checksum of the
3717 * resultant packet starting from the transport header-- the resultant checksum
3718 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3719 * header.
3720 */
3721static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3722{
76443456
AD
3723 unsigned char *csum_start = skb_transport_header(skb);
3724 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3725 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 3726
76443456
AD
3727 SKB_GSO_CB(skb)->csum = res;
3728 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 3729
76443456 3730 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
3731}
3732
bdcc0924 3733static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
3734{
3735 return skb_shinfo(skb)->gso_size;
3736}
3737
36a8f39e 3738/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 3739static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
3740{
3741 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3742}
3743
5293efe6
DB
3744static inline void skb_gso_reset(struct sk_buff *skb)
3745{
3746 skb_shinfo(skb)->gso_size = 0;
3747 skb_shinfo(skb)->gso_segs = 0;
3748 skb_shinfo(skb)->gso_type = 0;
3749}
3750
7965bd4d 3751void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
3752
3753static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3754{
3755 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3756 * wanted then gso_type will be set. */
05bdd2f1
ED
3757 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3758
b78462eb
AD
3759 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3760 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
3761 __skb_warn_lro_forwarding(skb);
3762 return true;
3763 }
3764 return false;
3765}
3766
35fc92a9
HX
3767static inline void skb_forward_csum(struct sk_buff *skb)
3768{
3769 /* Unfortunately we don't support this one. Any brave souls? */
3770 if (skb->ip_summed == CHECKSUM_COMPLETE)
3771 skb->ip_summed = CHECKSUM_NONE;
3772}
3773
bc8acf2c
ED
3774/**
3775 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3776 * @skb: skb to check
3777 *
3778 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3779 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3780 * use this helper, to document places where we make this assertion.
3781 */
05bdd2f1 3782static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
3783{
3784#ifdef DEBUG
3785 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3786#endif
3787}
3788
f35d9d8a 3789bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 3790
ed1f50c3 3791int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
3792struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3793 unsigned int transport_len,
3794 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 3795
3a7c1ee4
AD
3796/**
3797 * skb_head_is_locked - Determine if the skb->head is locked down
3798 * @skb: skb to check
3799 *
3800 * The head on skbs build around a head frag can be removed if they are
3801 * not cloned. This function returns true if the skb head is locked down
3802 * due to either being allocated via kmalloc, or by being a clone with
3803 * multiple references to the head.
3804 */
3805static inline bool skb_head_is_locked(const struct sk_buff *skb)
3806{
3807 return !skb->head_frag || skb_cloned(skb);
3808}
fe6cc55f
FW
3809
3810/**
3811 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3812 *
3813 * @skb: GSO skb
3814 *
3815 * skb_gso_network_seglen is used to determine the real size of the
3816 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3817 *
3818 * The MAC/L2 header is not accounted for.
3819 */
3820static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3821{
3822 unsigned int hdr_len = skb_transport_header(skb) -
3823 skb_network_header(skb);
3824 return hdr_len + skb_gso_transport_seglen(skb);
3825}
ee122c79 3826
179bc67f
EC
3827/* Local Checksum Offload.
3828 * Compute outer checksum based on the assumption that the
3829 * inner checksum will be offloaded later.
e8ae7b00
EC
3830 * See Documentation/networking/checksum-offloads.txt for
3831 * explanation of how this works.
179bc67f
EC
3832 * Fill in outer checksum adjustment (e.g. with sum of outer
3833 * pseudo-header) before calling.
3834 * Also ensure that inner checksum is in linear data area.
3835 */
3836static inline __wsum lco_csum(struct sk_buff *skb)
3837{
9e74a6da
AD
3838 unsigned char *csum_start = skb_checksum_start(skb);
3839 unsigned char *l4_hdr = skb_transport_header(skb);
3840 __wsum partial;
179bc67f
EC
3841
3842 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
3843 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3844 skb->csum_offset));
3845
179bc67f 3846 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 3847 * adjustment filled in by caller) and return result.
179bc67f 3848 */
9e74a6da 3849 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
3850}
3851
1da177e4
LT
3852#endif /* __KERNEL__ */
3853#endif /* _LINUX_SKBUFF_H */